Case 15 · Singapore · 22-person company, in-house legal team of 2 · Transformation: 6 weeks
Legal tech company, contract-lifecycle software vendor
A twenty-two-person contract-lifecycle software vendor whose own two-person legal team ran on the same ad hoc habits as the customers they sold software to - and whose customer-support team fielded AI questions from law firm buyers they were not equipped to answer credibly. Six weeks later the company's own legal function runs the MATTER Method internally, and the company can demonstrably say it uses its own operating model before selling anyone else on adopting theirs.
The challenge
The company built and sold contract-lifecycle software to law firms and legal departments, but its own two-person legal team - handling the company's customer contracts, vendor agreements, and a growing volume of AI-specific customer questions about data handling - ran on the same disorganised habits as many of the customers the product was built to serve. This was a credibility problem as much as an operational one: prospective law-firm customers evaluating the product asked pointed questions about the company's own AI governance and evidence-handling practices, and the honest answers were thinner than the product's marketing implied. Separately, the customer-support team was fielding an increasing volume of technical AI questions - about model choice, hallucination risk, data residency - from law firm buyers, without a consistent, accurate way to answer them.
How we diagnosed it
How we looked
- 01Read the company's own standard customer and vendor contracts, and interviewed the two-person legal team on how each was actually negotiated, tracked, and reviewed in practice.
- 02Sat in on four customer sales calls and three support tickets involving AI-related questions from prospective or existing law firm customers, to hear directly what was and was not being answered credibly.
- 03Reviewed the company's own AI product's marketing claims against its actual internal practices for handling customer data and model outputs.
- 04Interviewed the founding team on what 'using our own product's philosophy internally' would need to mean for it to be a credible claim rather than a talking point.
- 05Timed how long the two-person legal team actually took to turn around a standard vendor NDA against a matched pair of comparable counterparties, to establish a real baseline before proposing anything.
- 06Reviewed the company's own product documentation and customer-facing marketing materials against what the two-person legal team could actually demonstrate about its own AI governance practice when asked directly.
What we found
- ·The company's own contract review and negotiation process had no written playbook, no evidence-anchoring standard, and no matter container beyond a shared drive folder per counterparty - the exact pattern the company's own product existed to fix in its customers.
- ·Support staff answering AI-related buyer questions had no shared, accurate reference for the company's own model usage, data handling, or hallucination-mitigation practices, and answers varied by which support agent took the call.
- ·At least two recent sales calls had stalled specifically over unanswered or inconsistently answered questions about the company's own internal AI governance - a direct commercial cost traceable to the gap.
- ·The founding team had genuine, well-informed views on AI-native legal operations from building the product, but those views had never been applied to the company's own two-person legal function.
- ·Standard vendor NDA turnaround on the matched pair varied by more than double depending on which of the two legal team members handled it, with no written standard to explain the difference.
- ·Of the product marketing claims reviewed, several described governance practices - a documented review standard, a consistent AI-usage policy - that the two-person legal team could not actually produce evidence of when asked directly.
Primitives most affected
The transformation, phase by phase
01
Name
One week reading the company's own contracts and mapping its actual (undocumented) negotiation and review sequence. The bottleneck brief named the credibility gap explicitly - the company was selling an operating model it did not itself run - and the founding team agreed the internal build would double as the company's first fully honest answer to the AI-governance questions stalling sales conversations.
- Week 1: Read the company's own vendor and customer contracts, mapped the undocumented negotiation and review sequence, and delivered the bottleneck brief naming the credibility gap to the founding team.
02
Systematise
A matter container for the company's own contracts replaced the shared-drive folder, with the same required-field discipline adnah installs at any client. An evidence-anchoring rule was adopted for the legal team's own drafting and negotiation notes. A written, accurate internal reference on the company's actual AI usage, data handling, and model governance was built - the same document that support staff now use, verbatim, to answer buyer questions.
- Week 2: The matter container replaced the shared-drive folder for every active counterparty, and the evidence-anchoring rule was adopted for the legal team's own drafting and negotiation notes.
- Week 3: Drafted the written internal reference on the company's actual AI usage, data handling, and model-governance practice, checked line by line against what the team could actually demonstrate.
03
Automate
A drafting agent produces first-pass reviews of incoming vendor and customer contracts against the company's own now-named playbook. A support-triage agent routes AI-governance questions from prospective buyers to the accurate internal reference document automatically, with a named team member reviewing any question the reference does not clearly cover before a customer-facing answer goes out.
- Week 4: The internal drafting agent went live, producing first-pass reviews of incoming vendor and customer contracts against the newly named playbook.
- Week 5: The support-triage agent went live, routing AI-governance questions from prospective buyers to the accurate internal reference automatically.
- Week 6: Reviewed the first full week of both agents in production with the founding team and confirmed sales and support were using the same reference document consistently.
How we created value
A credible, honest answer to the question every serious buyer eventually asks
The company can now describe its own internal AI governance and evidence-handling practice accurately and specifically, because that practice now genuinely exists - closing the exact credibility gap that had stalled at least two recent sales conversations.
Consistent, accurate answers to AI questions regardless of which support agent takes the call
Support staff now route AI-governance questions to a single accurate reference rather than answering from individual understanding, which had previously varied noticeably from one agent to the next.
The company's own legal function became its best internal case study
Sales and marketing now have an accurate, first-party account of the MATTER Method running inside the company itself - not a hypothetical benefit described to prospects, but the company's own two-person legal team's actual working practice.
NDA turnaround stopped depending on which team member picked it up
The more-than-double spread in standard vendor NDA turnaround between the two legal team members closed once both worked from the same named playbook and the same first-pass agent output. Turnaround now depends on the counterparty, not on who happened to answer the request.
Marketing claims the legal team can now actually stand behind
The governance claims already in the company's own marketing that the legal team previously could not evidence when asked directly are now backed by a real practice - the gap between what was said publicly and what was true internally has closed rather than being quietly managed around.
AI education and training
Every deployment on this engagement was preceded by training - on what the models actually do, on OpenAI and Anthropic's behaviour on this organisation's own documents, and on the anchoring discipline required before automation, not after.
A founder-and-team session distinguishing the product's capability from the company's own discipline
The founding team, who understood AI capability deeply from building the product, worked through the distinction between a capable model and a disciplined practice around it - the same distinction the manifesto draws between AI-augmented and AI-native, applied to their own two-person legal function rather than to a customer's.
Support-team training on answering AI-governance questions accurately rather than persuasively
Support staff were trained specifically to answer buyer questions about model choice, hallucination risk, and data handling from the new accurate internal reference document - and, just as importantly, to say plainly when a question required escalation rather than to improvise a reassuring but inaccurate answer.
A worked comparison of OpenAI and Anthropic models on the company's own contract set
The two-person legal team ran the same contract-review task through both an OpenAI and an Anthropic model on the company's own vendor agreements, to reach a first-party, evidence-based view - rather than a marketing-informed one - on which performed better for which of the company's own document types.
A short session on the gap between what the marketing said and what was true
The founding team reviewed the specific marketing claims that outran the company's own internal practice, not to soften the claims but to close the gap - a direct, unflattering session that the founders themselves asked adnah to run rather than have discovered by a prospective customer first.
Smart solutions we built
A single-source-of-truth AI-governance reference, written once and used twice
The same document serves both as the internal governance record and, in an edited external-facing form, as the company's honest public answer to buyer questions about its own AI practices - eliminating the gap between internal reality and external claim that had been quietly costing sales.
A contract playbook the company's own product could eventually surface to customers
The named review sequence built during systematisation was structured deliberately in a format the company's own product team recognised as a candidate for a future customer-facing template feature - value created for the engagement and, potentially, for the product roadmap.
A vendor-NDA checklist that closed the turnaround gap between the two lawyers
One named, ordered checklist for standard vendor NDA review, built from the faster of the two team members' actual approach rather than an average of both, now governs both team members' work identically.
Agent solutions deployed
Internal contract drafting and review agent
Produces first-pass reviews of incoming vendor and customer contracts against the company's own newly named playbook, anchored and reviewed by the two-person legal team before anything is signed.
Support-triage agent
Routes AI-governance questions from prospective and existing customers to the accurate internal reference automatically, escalating anything the reference does not clearly cover to a named team member rather than allowing an improvised answer to reach a buyer.
NDA first-pass agent
Runs the standard vendor NDA checklist as a first pass against every incoming agreement, closing the turnaround gap between the two legal team members by giving both the same starting point.
Marketing-claim verification agent
Checks new marketing and sales copy mentioning the company's AI-governance practice against the current internal reference document before publication, flagging any claim the internal practice cannot currently support.
How MatterOS, LexOS, and the rest fit together
- MatterOS
- Runs the Matter and Team primitives for the company's own contracts - the matter container and the doer/reviewer role chart for the two-person legal team's own work, exactly as it would for any client.
- LexOS
- Runs the Evidence primitive - the anchoring rule applied to the company's own drafting and negotiation notes, and the structured internal reference document that both governs practice and answers customer questions.
- The company's own product
- Once the underlying process existed, the company's own contract-lifecycle software became a genuinely useful tool for its own team for the first time - previously under-used internally because there had been no named process for the product to support.
- Shared drive (retired)
- The per-counterparty shared drive folder that had served as the entire matter-tracking system was retired once the matter container went live, rather than kept as a fallback, so there was no ambiguity about which record was authoritative.
The shift to AI-native
This is the manifesto's argument turned back on the industry building the tools. A legal tech company has every structural advantage the thesis names - no leverage pyramid, no partnership vote, a founding team that already understands the technology - and yet this one had never applied its own stated philosophy to its own two-person legal function, because building a product and running an operating model are not the same discipline, and the company had only ever done the first. The commercial lesson is the sharpest part of this case: the credibility gap between what the company sold and what it practised was not an abstract governance concern, it was actively costing sales, and closing it turned the company's own legal team into first-party proof of the product's premise rather than a quiet embarrassment sitting behind it.
AI-native markers this firm now meets
- Pricing is transparent, not open-ended
- AI runs first, on every matter, systematically
- Evidence anchoring applied consistently
- Technical capability sits inside the organisation
Outcome
- Sales calls stalled on AI-governance questions
- → 0 since
- Internal contract review time
- −47%
- Support answers routed to accurate reference
- 0% → 100%
“We were selling the exact discipline we hadn't built for ourselves. Fixing that closed more sales conversations than any feature we shipped that quarter.”