IntelAgree vs Malbek: how they compare in 2026
IntelAgree and Malbek both sell contract lifecycle software to companies whose legal, sales, procurement and finance teams share one contract process, and both quote rather than publish a price. They differ on where the AI comes from and what it may learn. IntelAgree has built its own contract models since 2017, starting from contracts in SEC EDGAR filings. It sells its Saige Assist layer in five nested levels, from summaries in every plan up to playbook redlining in Word, extraction and risk scoring. It can accept a counterparty redline on its own when the change falls inside thresholds the customer sets. Malbek combines its own models with foundation models reached through AWS Bedrock and Azure, adds BusinessIQ for reading across the whole portfolio, and opens the repository to outside AI assistants through an MCP Server. Malbek's privacy policy rules out training on customer data, while IntelAgree says Saige learns from each customer's playbooks and corrections. IntelAgree suits a team that wants to buy AI level by level and let routine redlines clear on rules. Malbek suits a team that wants portfolio analytics, wide ERP connectors and a no training policy.
Which one fits
Choose IntelAgree if
- You want to buy AI in steps. Saige Assist comes in five nested levels, from comparison and summaries in every plan to chat, Word redlining, extraction and risk scoring, so a team pays only for the levels it uses.
- Counterparty redlines pile up. Saige classifies each one by intent and risk, accepts it automatically inside thresholds you set, and sends material changes to a person with the reason recorded.
- Signing authority must follow your HR system. Authority limits route each approval to the person empowered to sign, synced from systems such as Workday, and approval gates hold a contract when chosen terms appear.
- Staffing or CRM records start your contracts. Native connectors open contracts from Salesforce opportunities and from Bullhorn company, placement and candidate records, and Workato reaches more than 1,200 other apps.
- Each customer's data should sit apart. IntelAgree gives every subscriber a separate SQL Azure database with its own Key Vault keys, hosted in the Azure US East region.
Choose Malbek if
- Training on your contracts must be ruled out. Malbek's privacy policy says customer data does not train the AI models it uses, and its agreements with AI providers carry a clause against training.
- You want answers across the whole portfolio. BusinessIQ reads every contract for revenue, risk and obligations and answers through the Bek assistant with evidence from your own agreements.
- Outside AI assistants should work in the repository. The MCP Server lets them find, draft and negotiate contracts, keeping your permissions and logging every action on a timeline.
- Your ERP and procurement stack is broad. More than 50 connectors set up without code include NetSuite, SAP and SAP Ariba, Coupa, Workday, ServiceNow and a real time two way Salesforce sync that can serve several Salesforce orgs.
- Finance needs a SOC 1 report. Malbek holds SOC 1 Type II and SOC 2 Type II reports for 2026, released through its trust center on request.
At a glance
All 15 axes, side by side
The same grid applied to every vendor in the index, graded from public sources. Hover a grade to see what the letter means on that axis.
AI Centrality
How much of the product is actually AI. Whether the machine learning is the mechanism the buyer is paying for or a feature layered onto conventional software, and whether the vendor is specific about which is which.
IntelAgree describes itself as AI native since 2017, with extraction and metadata as the foundation and workflow built on top. The AI story page traces that history: models trained on contracts from SEC EDGAR filings in 2017, AutoML for customer trained extraction in 2018, contract classification for migrations in 2019, generative redlining against playbooks in 2021, contract question answering and risk scoring in 2022, configurable risk scores, approval gates and authority limits in 2025, and an agent with a prompt library in 2026. Saige Assist, the generative layer, runs in the app and inside Microsoft Word. The platform also delivers conventional CLM functions that would run without the models: templates, approval routing, eSignature, a repository, obligation tracking, renewals and integrations. Saige Assist is sold in levels, with comparison and summaries in every plan and the higher levels added as the customer chooses.
The models drive core capabilities on a platform that works without them. The contract lifecycle system underneath covers creation, workflow, approvals, the repository, milestones and clickwrap. The AI arrives as named modules. Malbek AI extracts terms, reviews third party paper against playbooks and runs the Bek assistant, AI Pro turns a customer's own templates into playbooks, and BusinessIQ reads across the whole portfolio. Modules are one of the four inputs a quote is built from, and the AI subprocessors apply only to customers who subscribe to AI services. The October 2026 Composable CLM release adds an MCP Server for outside agents and Drift Detective for customers with Bek and BusinessIQ. Remove the models and a working contract lifecycle system remains.
Citation Accuracy and Hallucination Disclosure
Whether the vendor publishes measured accuracy on citations and assertions, grounds output to primary sources, and says plainly what its system does when it does not know. Legal has a documented public record of fabricated citations reaching filed briefs, so an untested claim of accuracy is not evidence.
Saige Assist grounds its output in the customer's own material and shows where it came from. Every suggested edit traces to the customer's approved positions and clause library, a redline log records the clause source, and Saige Assist: Agent answers questions from the customer's contracts with a citation to the exact clause. The worked example on the Saige page shows a proposed liability clause pulled from the approved library with the edit tracked in Word. Answers draw on the customer's playbooks, precedent and approval rules rather than the open web. No accuracy figure, error rate, test set or evaluation is published for extraction, redlining, risk scoring or answers. The only figures near accuracy describe audit coverage: 100 percent audit traceability across AI assisted actions, on the legal solution page.
Grounding is described as an architecture. Bek's Ensemble LLM approach chains specialized agents that check one another's output, combining Malbek's own AI and machine learning models with foundation models reached through AWS Bedrock and Azure. Malbek states that this produces results that are well grounded and cited, and BusinessIQ answers through Bek with evidence drawn from the customer's contracts. The Malbek AI page claims unparalleled accuracy and reliability. No accuracy figure, test set or evaluation is published, and no failure mode is named, hallucination included. The source material is the customer's own repository, so grounding to primary law does not arise.
Autonomy and Oversight Model
What the system decides on its own, what a lawyer must approve, and whether the vendor documents where the review point sits. A tool that drafts under review and a tool that files without one are different products and different risks.
Some steps run on the AI alone, and in others a person decides. Saige Assist proposes language and legal accepts, edits or rejects each change, with every edit tracked in Word and tied to the playbook. On the legal solution page, Saige classifies each counterparty redline by intent and risk, accepts it automatically only within thresholds the customer has set, and sends material changes for human approval, with the reason and authority recorded. Approval gates hold a contract before signature when chosen terms appear, and authority limits route each approval to the person empowered to sign, synced from systems such as Workday. Saige Assist: Agent, an early release, asks for confirmation at every action. Risk scores use attributes, weights and bands the customer configures. IntelAgree's security FAQ adds that humans keep approval rights at every gate. Nothing published says what happens after an automatically accepted redline is found to be wrong.
In most features Malbek's AI recommends and a person decides. Playbook review flags deviations from preferred terms and recommends revisions, guided review gives each reviewer recommendations in real time, and Bek rewrites clauses that a user takes into the draft in Word. The privacy policy states that a score, rank or classification is a recommendation for a person to act on, which a user can review, override or ignore. The March 2025 Conversational Contracts release says customers can set verification steps so that critical decisions require human approval, while agents gather request details, start workflows and run bulk amendments. Actions taken through the MCP Server are logged with a timeline, and AI Pro playbooks sit under administrator control. A featured customer says about 80 percent of its templates no longer need legal review. No threshold at which an agent acts without a person is published, and nothing describes how a wrong output is corrected or reversed beyond the user's own review.
Operational and Outcome Evidence
Named, dated evidence that the product works in production at real firms or legal departments. Case studies with figures and identified customers count. Unattributed testimonials and launch announcements do not.
More than 25 named customer stories sit on the site. Headlines include Emmes cutting drafting time by nearly 85 percent, Everon saving 20 hours a week, Central Maine Healthcare saving $30,000 a year with eSignatures, Olsson reviewing routine contracts in about 15 minutes, Pebble Beach finalizing a contract in 9 minutes, GQR cutting a three day offer turnaround to seconds, Employbridge standardizing contracting across more than 300 branches, Enlyte making more than 10,000 contracts searchable and Spartan Race going live in 35 days. The home page quotes Eleanor Daugherty of MultiPlan and Casey Ralston of Command Alkon. The legal page states an annual renewal rate above 95 percent across enterprise customers. The story titles do not give deployment dates or say how the figures were measured, and most outcomes concern workflow and eSignature rather than the AI features.
Named customers and attributed quotes carry the evidence, without measurement. Customer logos include EDF, SurveyMonkey, Aptean, Articulate, Lighthouse, TIBCO, J.D. Power and Darden. Ellen Kranz, senior contracts specialist at Lighthouse, says about 80 percent of templates no longer need legal review and that contracts go active within minutes or hours. Adrienne Schaal, director of legal operations at Cloud Software Group, describes a Salesforce integration configured in minutes and field mapping without IT. Jeff Peduto, SVP of operations, describes renewal alerts 90 days ahead. The Salesforce page claims a 97 percent cut in legal touchpoints. A panel of headline figures, including 120 percent customer retention and 90 percent who would recommend Malbek, gives no basis, sample or date. Gartner named Malbek a Leader in its 2025 Magic Quadrant for contract life cycle management, and IDC named it a Major Player in its 2025 assessment of buy side contract applications.
Privilege and Confidentiality Posture
How client confidences are handled: attorney client privilege and work product treatment, segregation of one client matter from another, whether client data trains any model, and what the vendor commits to in writing rather than in marketing.
Customer data counts as Confidential Information under IntelAgree's published Terms of Service, to be used only for the purposes of the agreement and protected from third parties (section 7). The Data Protection Agreement treats personal data as the customer's confidential information and limits access to personnel who need it, and its security annex describes a separate SQL Azure database for each subscriber. The privacy policy, effective 21 July 2025, says the generative AI features do not share customer data with AI models accessible to third parties or the public. On the other side, the terms grant a perpetual license to customer data once aggregated and anonymized into Resultant Data, which IntelAgree owns and may publish, and Saige Assist is described as trained on the customer's playbooks and decision history. Privilege and work product are not addressed.
Malbek's commitments on training and confidentiality are substantive, short of privilege. The privacy policy says Malbek does not use customer data to train the AI models it uses, that data sent to an AI provider does not train that provider's models, that its provider agreements carry a clause against training, and that providers may not use prompts or outputs for their own purposes. It processes customer data only on the customer's documented instructions and does not use it for its own purposes. The trust center adds that inputs, outputs and embeddings are never shared and never accessible to the model vendor. Each tenant's data is logically separated, contract records can be restricted from viewing after signature, and retention and deletion are stated. The October 2026 release says Drift Detective benchmarks a draft against hundreds of thousands of customers' executed agreements, without saying whose agreements those are or how they are anonymized. These commitments sit in a policy and on the trust center, and the master subscription agreement is not public. The Terms of Use say no attorney client relationship or privilege is created with Malbek, and nothing addresses how privileged or work product material held in the repository is handled.
UPL and Professional Responsibility Posture
Whether the vendor is clear that it supplies a tool rather than legal advice, who its audience is, and how it addresses unauthorized practice of law, competence and supervision duties, and jurisdiction limits. ABA Formal Opinion 512 is the reference point. Where the advice line is not the duty a product raises, the axis is read through the nearest professional duty it does raise: judicial conduct rules and the reviewing duty for products sold only to courts, and the duty to bill for time actually spent for products that draft time entries.
No line between the product and legal advice is drawn in IntelAgree's terms, privacy policy, trust center or product pages, and none refers to lawyers' professional duties. The product is sold to procurement, sales, finance and business leaders as well as legal, and one Saige Assist level is named Contract Advice. The agent answers questions such as how to approach a renewal from the customer's own precedent, and Saige drafts redlines that favor the customer's side. The positions published are about control rather than advice: Saige proposes and legal approves, the expert stays in the loop, and approvals run through gates and authority limits the customer sets.
The Terms of Use draw the advice line. They state that Malbek is not a law firm or an attorney, may not perform services performed by an attorney and is not a substitute for one. Information on the service is not legal advice and is not guaranteed to be correct, complete or current, and no attorney client relationship or privilege is created. A user who believes Malbek has given legal advice about rights, remedies, defenses, forms or strategies is told not to proceed. The product is sold to legal, sales, procurement and finance teams inside companies, and the privacy policy treats scored or classified output as a recommendation for a person to act on. Nothing describes how the product supports a supervising lawyer's duties, and no jurisdiction limit is named beyond export controls. A featured customer says about 80 percent of its templates no longer need legal review.
AI Governance and Bias Disclosure
Published governance over model behavior: who owns it inside the vendor, what is tested before release, and what is disclosed about disparate output across matter types, parties, or populations.
Governed AI, the name IntelAgree gives its governance framework, centers on how a customer controls the AI: thresholds for automatic acceptance, approval gates, authority limits, configurable risk scores whose attribute weights and bands the customer sets, and an audit log of every user action and AI recommendation that exports to a SIEM. On its own side, it describes building training data from labeled SEC EDGAR contracts so it owns that data, and James Parks, its Chief Data Scientist, speaks publicly for its approach to where AI agents should and should not be used. No release testing process, test results or accountable owner for model behavior is published, and nothing addresses whether output differs across contract types, industries or languages.
Malbek's privacy policy, effective 3 September 2026, sets written rules for how its AI behaves. It names Malbek AI and BusinessIQ as the features that use AI, some of it on licensed third party models. It says the product tells users when they are dealing with an AI system, and marks generated or materially altered text as AI generated in machine readable form where that is feasible. Malbek makes no decisions with legal or similarly significant effects by automated means alone. A feature that scores, ranks or classifies a contract or a party produces a recommendation that a user can review, override or ignore, and Malbek commits to explain the system and offer human review if that changes. It also commits to advance notice before a new AI provider is added. The trust center's AI safeguards answer describes defenses against instruction override, data exfiltration through format manipulation, output hijacking, CSV injection, attack chains with several stages and delimiter confusion. Releases are certified by the QA team and opened through a management ticket. No one inside Malbek is named as accountable for AI outcomes, no testing of model output before release is described, and nothing is said about uneven output across contract types or counterparties.
AI Safety and Data Stewardship
Retention, deletion, access control, and what happens to prompts and documents after they are processed. Whether the vendor states its subprocessors and its incident practice, or leaves the buyer to assume.
Security controls in IntelAgree's trust center cover TLS 1.2 on every connection, SQL Transparent Data Encryption at rest, 256 bit AES file storage that is FIPS 140-2 compliant, secrets in Microsoft Key Vault, single sign on with SAML and OIDC, SCIM provisioning, role based access with custom roles, multifactor authentication on by default, an audit log of every action exportable to a SIEM, and point in time backups every 5 to 10 minutes with layered retention up to 10 years. The Data Protection Agreement commits to breach notice by email without undue delay, return and destruction of personal data within 30 days of a request or termination, a subprocessor list in its Annex III with notice of changes and 15 days to object, and a separate database per subscriber. The HIPAA agreement requires breach reports within five business days.
Retention, deletion, access, subprocessors and incident practice are each set out in Malbek's own documents. The privacy policy, effective 3 September 2026, keeps customer data for the term of the subscription and deletes or returns it at termination under the data processing addendum. Backups persist until their cycle is overwritten. Audit records cannot be altered and are kept for the life of the customer instance, so a deletion request removes the contract record and keeps only the log entry showing the deletion. The trust center adds that customers set retention periods for contracts, attachments and related data, that data is irreversibly deleted when a period ends, and that retention and deletion activity is logged. Access runs through role based permissions, single sign on, phishing resistant multifactor authentication and SCIM provisioning, and audit logs capture role, configuration and file changes. The subprocessor list names sixteen providers with function and location. If an incident affects customer data, Malbek notifies the customer without undue delay and within the period in the addendum, and supplies what the customer needs to meet the 72 hour deadline that applies under the GDPR. The addendum itself is released through the trust center on request.
AI Liability and Recourse
What the vendor stands behind contractually when its output is wrong. Indemnities, caps, carve outs, insurance, and whether any of it is published or only reachable through a negotiated agreement.
The Terms of Service published on IntelAgree's site are effective 17 December 2018 and name the provider as CoLabs, the company's earlier name, in square brackets. They provide the services as is and disclaim all implied warranties. The customer indemnifies the provider for its data and use of the services, and the terms contain no indemnity from the provider. Liability is capped at fees paid in the 12 months before the event, and the provider excludes liability for lost data and breaches of data or system security along with indirect damages. Florida law governs, with courts in Tampa. Nothing addresses loss from an AI redline, extraction or risk score that is wrong, and no insurance is published. Enterprise customers sign order forms whose terms are not published.
Malbek's Terms of Use, updated 9 March 2025, cover the website, its applications and plug ins, and the services provided through it. For a paying customer on a master subscription agreement, they apply only where that agreement says nothing different, and the master subscription agreement is not public. The Terms provide the service as is and with all faults. They disclaim every warranty, including accuracy, completeness and fitness for a particular purpose, and say Malbek does not warrant that the service is free of errors. Malbek's total liability is capped at the greater of $100 or the fees paid in the twelve months before the claim. The cap does not apply to death, property damage or personal injury caused by Malbek's gross negligence, or to injury caused by its fraud. The indemnity runs one way, from the customer to Malbek. The Terms give no indemnity for intellectual property or data claims, no remedy for a service that does not perform, and mention no insurance.
Practice Systems Integration Depth
How deeply the product reaches into the systems legal work already lives in: document management such as iManage and NetDocuments, Word and Outlook, contract lifecycle management, matter management, e-billing, and court filing systems.
Each integration has its own page covering what it does, the data flow, governance notes and evaluator questions. The Salesforce connector is native and two way: a rep starts a contract from an opportunity, counterparty and deal data fill in automatically, Saige Assist extracts terms and flags risk, and status and the signed contract mirror back to the Salesforce record, with Salesforce roles honored. Workday is a certified integration that connects contracts to supplier, vendor and worker data and can supply authority limits. Bullhorn contracts start from Company, Placement and Candidate records. Microsoft Dynamics 365, DocuSign, Adobe Acrobat Sign, HubSpot and Google Drive have their own connectors, Saige redlines inside Microsoft Word, and IntelAgree works in Outlook and Teams. Workato recipes reach more than 1,200 apps, and an open REST API with webhooks logs every sync event. No law firm document management system is listed.
Malbek documents integrations into the systems contracting runs through, and what they move. More than 50 are listed in a directory that filters by type, connection and direction, across CRM, ERP, electronic signature, storage, identity, collaboration and business intelligence. The Salesforce connector is native and syncs in both directions in real time. It covers accounts, opportunities, quotes, contacts, leads, products and custom objects, starts contract workflows from deal events, fills templates from opportunity data, and returns approval, signature, redlining and execution status to the opportunity record. One Malbek instance can connect several Salesforce orgs. Customers map fields through a drag and drop interface without code or IT. Named targets also include NetSuite, SAP and SAP Ariba, Coupa, Workday, Microsoft Dynamics, ServiceNow, Slack, Microsoft Word, SharePoint, DocuSign and Adobe Sign. A REST API covers systems without a connector, and the MCP Server opens the contract layer to outside AI assistants. The plans page says some integrations are included and others cost extra.
Deployment Model and Data Residency
Where the software runs and where the data sits. Multi tenant cloud, single tenant, private deployment, on premises, and whether region of residence is a published option or an enterprise conversation.
Microsoft Azure hosts IntelAgree, which is also listed on Microsoft Marketplace. The security annex to its Data Protection Agreement places all customer data in Azure's US East data center with active geo replication to US West, and provides a separate SQL Azure database for each subscriber, with an IP restricting firewall and Key Vault keys per subscription. For customers in the European Economic Area, transfers to the United States run under the Standard Contractual Clauses and the UK Addendum, with Portugal chosen for governing law and courts. No storage region outside the United States, private deployment or on premises option is published, and nothing states where AI processing runs as distinct from storage.
The tenancy model and the hosting country are both stated. The trust center describes a multitenant architecture in which each customer's contracts and data are logically separated, with security policies, patching and monitoring applied the same way across all tenants. Backups are encrypted with geographic redundancy, and the platform runs across several AWS availability zones. The privacy policy says production infrastructure is hosted in the United States and that personal data from other countries is transferred there, with standard contractual clauses for Europe, the United Kingdom and Switzerland. Some service providers operate in Ireland and Sweden. The AI services on the subprocessor list, AWS, Google Document AI and Microsoft Azure AI, are all listed in the United States. No choice of region and no dedicated single tenant option is described.
Security Certifications and Trust Center
Independent attestation a buyer can pull without a sales call: SOC 2, ISO 27001, penetration test summaries, a trust center with current reports and named scope rather than a badge image.
SOC 2 Type II examinations run each year by an independent third party against the Trust Services Criteria, with the full report available under NDA through a request form on IntelAgree's trust center. The security annex to its Data Protection Agreement adds an annual HIPAA Type 1 audit by a third party. HIPAA and GDPR compliance are stated, with a published business associate agreement and data protection agreement. The auditor, report period and scope are not named on the public pages, no ISO certification is listed, and the report is released on request rather than downloaded directly.
Malbek holds SOC 1 Type II and SOC 2 Type II attestations in its own name. Its trust center carries 2026 Type II reports for both, and the SOC 2 scope covers security, availability and confidentiality. The reports, a penetration testing summary, architecture and data flow diagrams and the data processing addendum are released through the trust center on request, under a confidentiality agreement where the document requires one. The ISO 9001, ISO 27001, ISO 27017, ISO 27018, PCI DSS Level 1 and SOC 1, 2 and 3 certifications listed beside them belong to AWS for the hosting, not to Malbek. Malbek also describes annual independent security audits, regular penetration testing, a security operations center, intrusion detection and quarterly security awareness training. No auditor or coverage period is published outside the reports.
Model Supply Chain Disclosure
Which models sit underneath, whose they are, where they run, and whether the vendor commits to telling customers when that changes. A legal buyer inherits every dependency it cannot see.
The AI is presented as IntelAgree's own: extraction models trained in house since 2017 on labeled contracts from SEC EDGAR filings, AutoML that lets customers train extraction for their own attributes, and Saige Assist as a contract specific generative layer with its own ontology and feedback loop rather than a generic model. The privacy policy, effective 21 July 2025, says the generative features do not share customer data with AI models accessible to third parties or the public. The subprocessor list in the Data Protection Agreement names Microsoft Azure for hosting and ABBYY for optical character recognition and names no model provider. Which foundation models sit under the generative features, and whether any is licensed from a third party, is not stated, and there is no commitment to notify customers of a model change.
Malbek names its AI providers but not its models. Malbek describes Bek's Ensemble LLM approach as combining its own AI and machine learning models with foundation models reached through AWS Bedrock and Azure. The subprocessor list names Amazon Web Services for hosting and AI services, and Google Document AI and Microsoft Azure AI for AI services, all in the United States, and marks the last two as applying only to customers on AI powered services. The privacy policy names the same three, commits to advance notice before a new AI provider is added under the data processing addendum, and says Malbek's agreements with AI providers carry a clause against training. Azure OpenAI also appears among the integration logos on the plans page. No individual model is named, the models offered through Bedrock are not identified, and no feature is mapped to a provider.
Commercial Transparency
Whether a buyer can learn what this costs without entering a sales process: published rates, the unit being charged, what sits behind an enterprise tier, and what implementation adds.
IntelAgree publishes no prices; its pricing path is a demo request and an ROI calculator. The package structure for the AI is published. Saige Assist comes in five nested levels, each including everything below it: Basic, with side by side comparison, summaries and help, in every plan; Chat, across a contract and up to 20 related agreements; Negotiation, with playbook redlining in Word; Data Extraction, with markup and bulk markup; and Risk Management, with risk scoring. IntelAgree presents this as paying only for capability a customer is ready to use. The 2018 Terms of Service set a one year initial term that renews month to month unless notice is given. Implementation is led by a dedicated team with a named account team, and its cost is not stated.
Malbek publishes the structure of its price without a figure. The plans page builds every quote from four inputs: user types, number of users, modules and Malbek connections, and says customers pay only for what they need. Some integrations are included and some cost extra. Implementation is scoped from a few data points and delivered in phases, until the customer can run and expand the platform on its own. Support comes by phone, email and the help center, with access to support and success teams after go live, and monthly training carries no extra cost. No number, band, plan name or feature split between packages is published, so the shape of the price is visible and its size is not.
Firm and Practice Coverage
Who the product is actually built for. AmLaw, midlaw, small firm and solo, in house departments, government and courts, and which practice areas are supported rather than merely claimed.
IntelAgree has solution pages for legal, procurement, sales and revenue operations, finance, administrators and business leaders, and industry pages for construction, financial services, healthcare, hospitality, insurance, manufacturing, retail, software, staffing, technology and transportation and logistics. It is built for both buying and selling contracts, with role specific views for procurement and revenue teams on one platform, and its healthcare offering includes a business associate agreement. Named customers range from staffing firms and health plans to construction, engineering and sports and leisure companies. Nothing published sets a minimum size, names contract types or languages the AI does not handle, or describes law firm use.
Malbek sells to corporate enterprises rather than law firms. Separate solution pages cover sales, legal, procurement and finance, each with its own purpose. Four industries also have their own pages: life sciences, manufacturing, consumer packaged goods and retail, and the life sciences page covers clinical, commercial and regulated agreements. Malbek states compliance with GDPR and CCPA. BusinessIQ's early adopters are Fortune Global 500 companies in healthcare and life sciences, technology and manufacturing, none of them named. Law firms are neither claimed nor excluded. No organization size is stated, government use is not addressed, and no contract type is named as unsupported.
The 12 legal signals, side by side
Recorded rather than graded. These are the questions a practitioner has to answer before a tool touches a client matter, and the answers are taken from public material only.
Client Data in Training
Can material a lawyer puts into this product be used to train a model?
Customer material trains the AI, by IntelAgree's own account. The Saige Assist page says the assistant is trained on the customer's playbooks and contract data rather than the open web and that every result teaches it, and the AI story page says that from 2020 every extraction a user corrected or confirmed made the whole platform smarter for everyone. The 2018 Terms of Service do not name training; they grant a perpetual license to customer data once aggregated and anonymized into Resultant Data, which IntelAgree owns and may use or publish in any lawful way.
The privacy policy, effective 21 July 2025, says the generative features do not share customer data with AI models accessible to third parties or the public. No opt out is described.
Malbek's policy says no, and its published Terms of Use never name training. The privacy policy, effective 3 September 2026, states that Malbek does not use customer data to train the AI models it uses. It adds that data sent to an AI provider does not train that provider's models, that Malbek's agreements with AI providers carry a clause against training, and that providers may not use prompts or outputs for their own purposes.
Malbek's trust center FAQ adds that data isolation means customer data is never used for model training. The Terms of Use let Malbek use customer data as needed to provide the service, and reserve a right to review content to operate and improve it, without naming training. The master subscription agreement and data processing addendum are not public. The October 2026 Composable CLM release says Drift Detective benchmarks a draft against hundreds of thousands of customers' executed agreements, and that the system keeps learning as more contracts are executed, without saying whose agreements those are.
Prompt and Output Retention
How long does the product keep what a lawyer typed, and can that be set to zero?
The Data Protection Agreement commits IntelAgree to return personal data on request and make reasonable efforts to destroy it within 30 days of the customer's written request or the end of the agreement, with exceptions for data it has a legal right or duty to keep. The 2018 Terms of Service add return or destruction of customer data on written request within 30 days of termination. The trust center describes point in time backups every 5 to 10 minutes with layered retention up to 10 years, and an audit log of every user action and AI recommendation. No separate retention period is stated for Saige Assist prompts, answers or redline suggestions.
The customer sets the retention window for its contracts, and no separate window is stated for prompts. Customers define retention periods for contracts, attachments and related data, data is irreversibly deleted when a period ends, and retention and deletion activity is logged. The privacy policy keeps customer data for the term of the subscription and deletes or returns it at termination under the data processing addendum, with backups kept until their cycle is overwritten.
Audit records cannot be altered and are kept for the life of the customer instance. AI providers may not use prompts or outputs for their own purposes, but no period is given for how long prompts and Bek conversations are kept. No setting to keep nothing is offered.
Ethical Walls and Matter Segregation
Does retrieval respect the firm’s ethical walls, or can the model read across them?
Between customers, the security annex to the Data Protection Agreement provides a separate SQL Azure database for each subscriber, with Key Vault keys per subscription and a firewall that can restrict access by IP address. Within a customer, IntelAgree runs its own role based access control with custom roles, SCIM provisioning and multifactor authentication on by default, plus approval gates and authority limits that decide who signs off on what, and the Salesforce integration honors Salesforce roles.
Procurement and revenue teams work on one platform with views and controls set by role, and out of office coverage hands approvals to audited stand ins. The Saige page shows redline logs marked as permission gated. No published guide sets out the permission levels or how Saige Assist: Agent applies them when it searches across agreements.
Malbek runs its own permission model, and the customer keeps it aligned. Role based access gives each user only the contracts, data and features their role needs, and SCIM provisioning from the identity provider keeps accounts and roles in step with the directory. Contract records can be restricted from viewing or use after signature, under the customer administrator's control, and permission changes are captured by the audit function.
Each tenant's data is logically separated in a multitenant architecture. The Malbek AI page says the MCP Server preserves permissions, business rules and an audit record when outside agents act. Nothing describes walls between matters or business units inside one customer's instance, or how Bek and BusinessIQ answers respect a user's permissions.
Third Party Request and Subpoena Notice
If someone subpoenas the vendor for a firm’s data, does the firm hear about it first?
Exhibit B to the Data Protection Agreement commits IntelAgree to notify the customer promptly of any request for personal data from a government, regulator or law enforcement authority unless law or a binding order forbids it, to help the customer respond, to seek the customer's consent before any voluntary disclosure of unencrypted personal data, and to keep an internal procedure for such requests. It also states that no back doors have been built for government access.
The privacy policy reserves disclosure to meet a subpoena, legal process or an enforceable governmental request. No transparency report is published.
Malbek's privacy policy commits to telling the customer first. When Malbek receives a demand for customer data, it reviews the demand for validity and scope, challenges demands that are overbroad or unlawful, and redirects the requester to the customer where the customer is the right party to respond. It notifies the affected customer before disclosing anything unless the law prohibits it, and then as soon as the prohibition lifts, and discloses no more than the demand requires.
The Terms of Use separately allow Malbek to share information to comply with a subpoena or other legal process and say nothing about notice. The master subscription agreement is not public. No transparency report is published.
Primary Law Corpus Provenance
Where does the law in this product come from, and does the vendor have the right to use it?
The first models came from public SEC EDGAR filings: in 2017 IntelAgree pulled contracts from them and labeled them by hand to train extraction of clauses such as renewal dates, governing law and indemnity, which it says gave it ownership of that training data. Saige Assist works from the customer's own playbooks, templates, clause library and decision history. The product does not retrieve case law or legislation. Nothing describes later training sets, any licensed data or the data behind any foundation model under the generative features.
No outside legal content feeds Malbek's answers. The product analyzes the customer's own contract repository, and BusinessIQ distills that repository into intelligence on revenue, risk and obligations. Drift Detective benchmarks drafts against executed customer agreements rather than a legal publisher's content. No external database, publisher, content license or jurisdiction coverage is named.
Good Law Verification
Does the product tell you when the authority it just cited has been overruled?
IntelAgree manages a customer's own contracts and does not cite case law or legislation, so a citator is not part of the product. Nothing on the Saige Assist, AI story, legal or trust center pages addresses checking authority for later history. The agent page says the questions a team asks of its contracts change when a new tariff, regulation or force majeure event arrives, and nothing describes the product checking playbook positions against changes in law.
Malbek does not retrieve or cite primary law. It works on the customer's own contracts, surfacing clause level risk, obligations and commercial exposure rather than case law or statute. Nothing on its pages addresses checking an authority for subsequent history, which this kind of product has no reason to do.
Refusal and Uncertainty Behavior
What does the product do when the answer is not in the corpus?
No path for declining to answer is documented, and no confidence score is shown for Saige Assist answers or extractions. The routing that is published runs on risk rather than confidence: counterparty redlines are classified by intent and risk, accepted automatically within thresholds the customer sets, and sent to a person when material, and the agent asks for confirmation before each action. Nothing says what the agent does when the customer's contracts do not answer a question.
Agents check one another's work, and no way for the product to decline is described. Bek's Ensemble LLM approach chains specialized agents that cross validate output, which is a quality control rather than a refusal path. No confidence score, grounding score or answer of no support is described as visible to the user. The trust center's AI safeguards answer deals with attacks on the model rather than model error, and hallucination is not named on Malbek's pages.
Fabricated Citation Record
Does a public court record exist addressing fabricated or hallucinated legal citations in output from this product?
The AI Hallucination Cases database maintained by Damien Charlotin, which records court decisions worldwide that address hallucinated AI content and the tool involved where known, has no entry naming IntelAgree, Saige Assist or CoLabs. This is a statement about the public record rather than a finding about the product, and it covers fabricated content only. IntelAgree manages commercial contracts rather than producing court filings, so its output does not ordinarily reach a brief.
No court order, opinion or disciplinary record in the AI Hallucination Cases database names Malbek or its Bek assistant. Malbek's AI drafts, reviews and summarizes contract language rather than citing case law, and its output still needs a person to check it. This is a statement about the public record on one subject, not a finding about the product.
Bar Guidance Alignment
Has the vendor engaged in public with the ethics opinions its buyers are bound by?
No bar opinion, ethics rule, statement that the product does not give legal advice, or other reference to lawyers' professional obligations appears in IntelAgree's Terms of Service, privacy policy, Data Protection Agreement, trust center or Saige Assist and legal pages. The compliance positions published are HIPAA, GDPR and SOC 2, which bind IntelAgree as a provider. A general counsel, Lee Rone, is named on the team page as responsible for IntelAgree's own contracts and as an advisor on product design.
Ethics, in Malbek's materials, means corporate conduct rather than professional duty. The trust center's ethics answer covers policies against slavery, bribery and discrimination, and on sustainability. No bar association, ethics opinion or professional code is named, and the Terms of Use draw the advice line without citing any guidance. The buyer is a corporate department rather than a regulated practitioner.
Billing and Fee Posture
Does the vendor address what happens to the bill when the work takes an hour instead of six?
IntelAgree is bought by companies to run their own buying and selling contracts, and its named customers are staffing firms, health plans, construction and engineering companies, retailers and leisure businesses rather than law firms, so no client bill sits in the loop. Its savings claims are aimed at the buyer's own time and cost, such as Emmes cutting drafting time by nearly 85 percent and Everon saving 20 hours a week. Nothing published addresses fee treatment of AI assisted work for a firm that bills a client.
The product sits outside a fee between a lawyer and a client. Malbek is bought by corporate legal, sales, procurement and finance teams, which bill no client for contract work. Its savings claims aim at the buyer's own time and cost. A featured customer says about 80 percent of templates no longer need legal review. Malbek claims a 97 percent cut in legal touchpoints through its Salesforce connector, and that 91 percent of customers report higher productivity. No law firm is a named buyer segment.
Outside Counsel Guideline Readiness
Can a firm get this vendor through a client’s AI clause without a bespoke negotiation?
Annex III to the published Data Protection Agreement lists IntelAgree's subprocessors: Microsoft Azure for hosting, ABBYY for optical character recognition, SocketLabs for email, Elevate for systems integration, WalkMe for digital adoption, Cloud Factory and Epiq for data migration, Citrix for file transfer, Gong for recorded sessions, and Waypoint, Cimplifi and Koho Consulting for implementation. The list names no AI model provider.
The Data Protection Agreement and a HIPAA business associate agreement are published as templates, the latter with breach reports within five business days of discovery and Florida governing law, and the SOC 2 Type II report is released under NDA on request. No client facing AI disclosure material is published.
Malbek publishes a current subprocessor list on its trust center, with sixteen providers and the function and location of each. Amazon Web Services carries hosting and AI services, and Google Document AI and Microsoft Azure AI are AI services that apply only to customers on AI powered features. The privacy policy names the same AI subprocessors, commits to advance notice before a new AI provider is added, and says Malbek's agreements with those providers forbid training.
The trust center FAQ adds that foundation models are reached through AWS Bedrock and Azure. The data processing addendum, SOC reports and architecture documents are listed on the trust center and released on request, under a confidentiality agreement where needed.
Court Disclosure Support
If a judge’s standing order requires an AI disclosure, can the product produce one?
IntelAgree keeps an audit log of every user action and AI recommendation, exportable to a customer's SIEM, and its legal page says the full provenance chain for an AI assisted action, covering the clause source, the approver, the context and the threshold applied, exports on demand, with 100 percent audit traceability across AI assisted actions. Saige redlines are tracked in Word and tied to the playbook position they came from.
Nothing published says the record names the model that produced a suggestion, and no disclosure template is published. The output is contracts and contract records rather than court filings.
Parts of a disclosure record exist, with no export built for one. The privacy policy says text the service generates or materially alters is marked as AI generated in machine readable form, so far as that is technically feasible. Audit logs capture role, configuration and file changes, and the October 2026 release says every action taken through the MCP Server is logged with a timeline. No export of the model used, the sources retrieved and the person who checked an output is described. The product serves corporate contracting rather than litigation, so no court sits in its path.
The questions both sides leave open
Derived from the records above rather than written, so it cannot favor either vendor. Take these into both conversations and ask each side the same question.
- Good Law Verification
- Refusal and Uncertainty Behavior
- Bar Guidance Alignment
Product updates
The latest logged changes from each vendor, newest first, each linked to its source.
Malbek
-
Integration / interoperability
Malbek made its MCP Server generally available at its Envision 2026 event, as part of what it calls Composable CLM.
Source
In summary
IntelAgree
IntelAgree, founded in Tampa, is an enterprise contract lifecycle platform for both buying and selling contracts, with drafting, approvals, eSignature, a repository, obligation tracking and renewals. Its Saige Assist layer compares and summarizes contracts in every plan and answers questions across a contract and up to 20 related agreements. Higher levels redline in Word against the customer's playbook, extract data from a plain description and score risk on attributes the customer weights. Named customers include Emmes, which cut drafting time by nearly 85 percent, and Enlyte, which made more than 10,000 contracts searchable. According to the AI Legal Index, IntelAgree runs on Microsoft Azure in the United States with a separate database for each customer and holds a SOC 2 Type II report released under NDA. The main caveat is data use: Saige learns from customer playbooks and corrections, and the 2018 terms grant IntelAgree rights to customer data once anonymized and aggregated.
Malbek
Malbek, headquartered in Princeton, New Jersey, is a contract lifecycle platform for corporate legal departments and the sales, procurement and finance teams beside them. Malbek AI extracts terms and reviews third party paper against playbooks, and the Bek assistant answers questions by text or voice and rewrites clauses in Word. BusinessIQ, generally available since April 2026, reads the whole portfolio for revenue, risk and obligations. The October 2026 Composable CLM release added an MCP Server for outside AI assistants and Drift Detective for benchmarking drafts. According to the AI Legal Index, Malbek's privacy policy bars training on customer data, production runs on AWS in the United States, and the company holds SOC 1 Type II and SOC 2 Type II reports. The main caveat is the paper: the master subscription agreement is not public, and the public Terms of Use cap liability at the greater of $100 or twelve months of fees.
Questions buyers ask
How does IntelAgree compare to Malbek?
Both are contract lifecycle platforms for companies whose legal, sales, procurement and finance teams share one contract process. IntelAgree's AI is its own contract models, sold as five Saige Assist levels, with automatic acceptance of redlines inside set thresholds. Malbek blends its own models with foundation models through AWS Bedrock and Azure, and adds BusinessIQ portfolio analytics and an MCP Server for outside AI assistants. From the AI Legal Index, based on each vendor's own published materials as of November 16, 2026. No vendor pays for placement.
What do IntelAgree and Malbek cost?
Neither publishes a figure, and both quote. IntelAgree prices Saige Assist by level, with comparison and summaries in every plan, and its 2018 terms set a one year initial term that then renews month to month. Malbek builds each quote from user types, number of users, modules and connections, with some integrations included and others extra, and monthly training at no added cost. From the AI Legal Index, based on each vendor's own published materials as of November 16, 2026. No vendor pays for placement.
Do IntelAgree and Malbek train AI on customer contracts?
Malbek's privacy policy says customer data does not train the models it uses, and its AI provider agreements forbid training. IntelAgree says Saige Assist is trained on each customer's playbooks and contract data and learns from every result, and its terms grant it a perpetual license to customer data once anonymized and aggregated. IntelAgree's privacy policy adds that its generative features do not share customer data with AI models open to third parties. From the AI Legal Index, based on each vendor's own published materials as of November 16, 2026. No vendor pays for placement.
Which AI models do IntelAgree and Malbek use?
IntelAgree presents its AI as its own, with extraction models trained in house since 2017 on labeled SEC EDGAR contracts, and its subprocessor list names Microsoft Azure and ABBYY but no model provider. Malbek reaches foundation models through AWS Bedrock and Azure in what it calls an Ensemble LLM approach, and lists AWS, Google Document AI and Microsoft Azure AI as AI subprocessors. Neither names an individual model. From the AI Legal Index, based on each vendor's own published materials as of November 16, 2026. No vendor pays for placement.
Where do IntelAgree and Malbek host contract data?
IntelAgree hosts on Microsoft Azure in US East with geo replication to US West and a separate database for each subscriber. Malbek runs production on AWS in the United States across several availability zones, with each tenant's data logically separated and no region choice. Both move European data to the United States under standard contractual clauses. From the AI Legal Index, based on each vendor's own published materials as of November 16, 2026. No vendor pays for placement.
Neither vendor paid for inclusion, placement or a grade, and neither reviewed this page before it published. Everything on it comes from public material on the dates shown. How the index grades.
Neither vendor publishes an accuracy figure or test set for extraction, review or answers, and neither names an individual model behind its generative features. IntelAgree's public terms date from 2018 under its earlier name, CoLabs. They give no indemnity from the provider and say nothing about loss from a wrong AI redline or an automatically accepted change. Malbek's master subscription agreement and data processing addendum are not public. Malbek does not say whose executed agreements Drift Detective benchmarks against, or how long Bek conversations are kept.