We’ll build it — under our own scorecard.
Available by exception, after we’ve engaged on red-team. Same scorecard discipline we’d use to grade a third party, applied throughout the build. Independence is structurally enforced — the build lead and the red-team lead on your engagement are different people, named in your contract. Capped at 30% of our annual revenue, on purpose.
Methodology
/ 4 phasesEligibility
We will only build for clients with whom we have already engaged on red-team. Outside-in build engagements get declined. The red-team relationship is the qualifier.
Independence-clause review
Your legal counsel and ours sign the independence schedule before any build work starts. Named build lead, named red-team lead, no overlap. The schedule names the dispute path if the red-team lead vetoes a build decision.
Build to the scorecard
The build team scores against the same suite a third-party auditor would use against the result. The suite is published before the agent ships, not after.
Public scorecard
Every build engagement ends with a public scorecard at lattice.ai/evals, signed by the red-team lead — not the build lead. If the scorecard would embarrass either side, the engagement doesn’t conclude there.
Engagement shape
/ what you sign up for- 01The agent — built against your domain, scored continuously
- 02Eval suite — written before the agent, published at handover
- 03Public scorecard at handover — signed by the red-team lead, not the build lead
- 04Annual revenue cap — engagement contributes against our ≤ 30% build-revenue ceiling
Mappings
/ 3 frameworksEvidence
/ forthcomingThe proof for this pillar gets linked here as we ship public scorecards and clear case studies for publication. We don’t backfill this section with placeholders — when evidence lands, it lands here.
Why the cap exists
If we ever derive more than 30% of annual revenue from build, the firm’s incentives drift. The cap is in the company’s articles, not just the website. We publish the percentage in our annual report.
The cap is also why we say no to build engagements every quarter. It’s not a flex. It’s a structural choice that makes the red-team work credible.
Why we build at all
Some clients ask us to build because they want the evidence-based discipline applied throughout the engineering, not just at the end. We accept some of those. Not most.
The shape that works: the client wants someone to build the eval suite, run the build against it, and ship the scorecard publicly. The shape that doesn’t work: the client wants a build partner and is sourcing evidence-based assurance as a side benefit. We refer those to other firms.