Luminance

Luminance is a contract-lifecycle product aimed at work that starts with a draft agreement and continues through negotiation and contract intelligence. Its public negotiation page describes identifying non-standard clauses, proposing markups against a company's gold standards, using checklists in Microsoft Word, and offering alternatives informed by templates, playbooks, and prior negotiations. [S1]

The January 2026 product release describes negotiation AI, workflow orchestration, portfolio-wide contract intelligence, and a question-answering assistant with cited answers. These are vendor statements about a recently released platform; a buyer should confirm actual availability and test the specific workflow, model behavior, authority scope, and integration configuration it is buying. [S2]

Luminance is a plausible fit for a legal team that already has settled positions for a routine agreement and can hold legal owners accountable for final advice and exceptions. It is not a substitute for legal judgment: the public trial terms expressly say that its outputs and functionality are not legal advice or a substitute for independent or internal legal advice. [S4]

Best for: legal and contract-operations teams with recurring agreements, approved standards, named legal owners, and a controlled Microsoft Word-centric pilot.

Not for: a buyer seeking unsupervised legal advice, automatic approval of novel or high-risk agreements, or broad contract-system writes before its data, permissions, escalation, audit, and rollback controls are validated.

Triage an incoming agreement against approved positions

• Trigger: A routine counterparty agreement arrives for a defined first review.
• Inputs: Agreement text, buyer-approved gold standards, templates, playbooks, prior negotiation context where enabled, and the responsible team's instructions.
• Output: Identified non-standard or risky language, proposed markups or alternatives, and a checklist for negotiation positions.
• Human checkpoint: The legal owner checks the agreement, business context, suggested change, and relevant approvals before any position is sent or accepted.
• Success measures: time to first pass, material-issue recall versus an independent legal review, false-positive rate, accepted and rejected suggestions, rework, exception volume, and time to signature.

Luminance says its negotiation product can mark up risky or non-standard clauses against a company's gold standards, spot departures from standard wording, and provide alternatives based on prior negotiations, templates, and playbooks. Its page places checklists in Microsoft Word. Treat this as a capability claim to test with the buyer's own language and counterparties. [S1]

Focus legal attention on exceptions

• Trigger: A reviewer finds a departure from the team's standard terms or an agreement that needs commercial, privacy, security, or executive input.
• Inputs: The agreement, the identified departure, approved fallbacks, relevant internal policy, counterparty context, and the buyer's escalation rules.
• Output: A documented exception, proposed response, approval request, or revised position.
• Human checkpoint: The designated approver decides whether to accept risk, revise the position, seek further legal analysis, or stop the deal.
• Success measures: complete routing of mandatory exceptions, approval completeness, unresolved-exception age, unauthorized-change rate, and final position consistency.

The public site describes checklist-guided attention to core negotiating positions and alternatives grounded in business context. It does not establish the buyer's exact routing, approval, or integration behavior, so those need a controlled test. [S1]

Query the contract portfolio before acting

• Trigger: Legal or an authorized business partner needs a focused answer about a contract, amendment, obligation, or negotiation history.
• Inputs: Permitted contracts and amendments, the approved user role, a narrowly phrased question, and the buyer's source-access rules.
• Output: A candidate answer with source material or citations for the reviewer to inspect.
• Human checkpoint: An authorized reviewer checks the underlying contract and cited material, then decides whether it supports the proposed action or needs further legal analysis.
• Success measures: correct-source retrieval, citation usefulness, answer accuracy on a reviewed sample, access-control errors, search time, and escalations.

Luminance's January 2026 release says its contract-intelligence capability can query contracts, families, amendments, and obligations, and that Ask Lumi produces cited answers. Buyers should verify the feature's purchased availability, data boundary, and citation quality with their own portfolio. [S2]

The conservative operating model is: authorized user selects an agreement and allowed context → Luminance highlights departures and proposes a response → legal owner verifies the source agreement, business facts, and decision → required commercial or risk approvers resolve exceptions → the buyer's normal execution process sends, signs, and records the agreement. The public materials support the first review and contract-intelligence parts of this model; the approval and execution controls must be supplied and tested by the buyer. [S1][S2]

Luminance also markets autonomous negotiation. Its public page says that automated negotiation can review agreements, apply legal standards, redline terms, send revisions, track counterparty responses, and explain reasoning. This makes authority boundaries a primary procurement issue, not a reason to reduce oversight. Do not activate that scope until a buyer has tested representative and adversarial contracts, defined escalation thresholds, and verified how to stop or reverse activity. [S1]

Verified facts from primary sources

• Luminance's negotiation page describes marking up risky or non-standard clauses against gold standards, identifying departures from standard wording, providing alternatives based on templates, playbooks, and prior negotiations, and using checklists in Microsoft Word. [S1]
• In a January 27, 2026 release, Luminance says its updated platform includes negotiation AI, workflow orchestration, contract intelligence across contracts and obligations, and Ask Lumi with cited answers; it says broader availability began February 25 after beta use with design partners. [S2]
• Luminance's trial terms, last updated April 1, 2025, describe a proof-of-value trial that is normally two weeks and say the product is hosted in an AWS Region chosen by Luminance unless otherwise agreed. [S4]
• The public security page states that Luminance is certified to ISO 27001:2022, has completed a SOC 2 Type 2 examination, uses dedicated single-tenant AWS instances, encrypts data at rest and in transit, and provides described role, MFA, audit, and division-level permission controls. [S3]
• The public privacy policy, last amended July 27, 2023, says an existing business customer's contract governs the processing relationship and points to applicable product or service descriptions for additional processing details. [S5]

Vendor claims that need buyer validation

• Luminance says its platform can automate and improve contract work and offers substantial time savings. These claims are vendor statements, not a forecast for another team's contracts, users, or approval model. [S2]
• Luminance says every answer in the new platform is grounded in source material with visible citations. Test source coverage, citation correctness, and behavior when agreements are incomplete, conflicting, or outside the expected scope. [S2]
• Luminance says its security practices include a completed SOC 2 Type 2 examination and independent penetration testing. Obtain the current reports, scope, exceptions, bridge period, applicable entity, deployment configuration, and contractual commitments rather than relying on the public summary. [S3]

B2Bagents assessment

Luminance should be evaluated as a controlled contract-review and negotiation platform that can progress from legal copilot work to more automated activity. That assessment follows from the public product workflow and its stated autonomous-negotiation capability, balanced against its own terms saying the service is not legal advice. The deciding evidence is not a general productivity claim: it is whether the buyer can show that suggested and automated actions follow its standards, reach the right approver, preserve privilege and access boundaries, and can be stopped or corrected. [S1][S4]

Material unknowns

• Public subscription price, price metric, minimum commitment, module availability, implementation scope, support level, usage limits, renewal, cancellation, export, and termination assistance.
• Exact Word, Microsoft 365, repository, API, e-signature, SSO, role, audit, export, admin, and error-handling behavior for the buyer's purchased edition and tenant.
• Current model-provider and subprocessor list, DPA, product-specific retention, data residency, customer support access, legal-hold treatment, deletion verification, backups, incident commitments, and the scope of any data used for learning.
• The real accuracy, material-issue recall, false-positive rate, citation quality, jurisdiction coverage, playbook fidelity, and performance on the buyer's contracts and counterparty positions.
• The authority threshold, human escalation, audit evidence, outbound communication controls, emergency stop, and rollback path for autonomous negotiation or workflow writes.

Begin with one high-volume, low-risk agreement type for which the legal team has approved fallback language and a named owner. Use a non-production or otherwise authorized document set. Define allowed data sources, user roles, writing authority, approvals, escalation, and shutdown before connecting wider repositories or enabling automated outbound activity.

Set a baseline for first-pass and final-approval time, lawyer hours, independently reviewed material-issue recall, false positives, missed clauses, redline acceptance and rework, exception rate, approval completeness, retrieval accuracy, access incidents, and time to recover from a bad change. Include difficult cases: conflicting precedent, ambiguous business context, missing schedules, unusual limitation or indemnity language, multiple jurisdictions, tracked changes, sensitive documents, revoked roles, malformed source files, and failed integrations.

Agree success thresholds before the pilot: faster first pass without a material decline in independently reviewed issue coverage; correct application of the team's playbook; source material available for each material recommendation; complete legal-owner review for final positions; no unauthorized access or outbound action; and a tested pause, export, correction, and rollback route. Stop for a material missed issue, unreliable citation, unexplained output, access-control failure, untraceable approval, unexpected data exposure, or an automation boundary that cannot be safely governed.

• Spellbook: compare Luminance's broader contract-lifecycle and workflow position with a Word-centered contract-review product and the legal team's required playbook controls.
• Harvey: compare contract-focused workflow automation with a broader legal AI environment, especially on source grounding, data boundaries, and lawyer review.
• Crosby: compare buyer-operated contract software with other transactional legal workflows on integration depth, authority boundaries, and commercial model.

1. Which exact product modules and autonomous-negotiation functions are available in our purchased edition, and what user roles may trigger a redline, send a revision, or change a workflow state?
2. For every enabled source and integration, which documents, prompts, metadata, outputs, and logs are read, written, retained, backed up, exported, deleted, or available to support personnel and subprocessors?
3. Can the vendor demonstrate citation and reasoning behavior on our agreement types, jurisdictions, missing facts, conflicting standards, unusual counterparties, and adversarial language?
4. What independent evidence covers our deployed service: ISO 27001 scope, SOC 2 Type 2 report, recent bridge letter, pen-test scope, DPA, model and subprocessor list, data residency, incident notice, and assurance exceptions?
5. How do SSO, MFA, RBAC, division permissions, offboarding, legal holds, audit logs, emergency stop, export, correction, and rollback work during a role change or integration failure?
6. What will implementation, training, support, proof of value, subscription, usage, renewal, termination, and migration cost for the pilot and an expanded deployment?