Before you write
the brief.
The questions we get asked most, answered plainly: no hedging, no marketing copy.
01 How do engagements usually start?
With a discovery session, not a sales call. We map the problem, the data, and the constraints, then send a written brief and estimate. If the scope is still fuzzy after that, we'll sometimes run a small paid discovery phase first rather than guess. See the Process page for the full shape of an engagement.
Link to this answer: How do engagements usually start?02 How much does this cost, and how long does it take?
It depends on scope, so we quote ranges, not a fixed price, once we understand the problem. Most first engagements ship a production slice within a few weeks. The budget ranges on the contact form are a reasonable starting point for sizing your brief.
Link to this answer: How much does this cost, and how long does it take?03 Who owns the code and the models?
You do. Everything we build is documented and observable. We hand it over at the end of the engagement. We're consultants, not vendors with a kill-switch.
Link to this answer: Who owns the code and the models?04 Do you use our data to train models?
No. Your data stays in your environment. We build on top of it, not off it.
Link to this answer: Do you use our data to train models?05 We're not sure AI is even the right answer. Will you tell us?
Yes. Our Strategy & Readiness engagements exist specifically to give you an honest answer, including "don't build this" when that's the right call.
Link to this answer: We're not sure AI is even the right answer. Will you tell us?06 Which models or stack do you build on?
We're model-agnostic — Anthropic, OpenAI, open-weight, or self-hosted — and we build on open standards rather than locking you into one vendor's stack.
Link to this answer: Which models or stack do you build on?Every other question, answered on the page that owns it. #
The six questions above come up before a brief. Every other question the site answers already has a home on a real page. Search below, or browse by group.
Showing all 206 questions
Before you start
What it costs, and who ends up owning the result.
What we build
Our products, and the architecture and tooling underneath them.
Context Engineering
- Isn't this just prompt engineering with a longer name?
- How is this different from /tool-design/?
- We just bought a model with a much bigger context window. Does this still matter?
- What's the single most common mistake on this page?
- Do we need all 24 practices before our first agent ships?
- Where does the memory surface stop and a vector database start?
MCP
- Isn't MCP itself the security boundary?
- We only connect first-party servers we wrote ourselves. Does the trust boundary still apply?
- What's the single fastest way to get burned on this surface?
- Do we need all 24 controls before connecting a single server?
- How is this different from the trust-boundary surface on /tool-design/?
- Does Dynamic Client Registration still work?
Running it safely
Governance and security review, plus the adoption and release controls that keep a system honest.
Adoption
- Isn't this just change management with new labels?
- We already ran a successful pilot. Why isn't that enough?
- What's the fastest way to kill a rollout without meaning to?
- How do we know adoption is real, rather than seat activation with nothing behind it?
- Who should own this, the platform team or engineering leadership?
- Does this still apply once we're past the pilot and scaling org-wide?
Governance
- We're not building a high-risk system. Do we need any of this?
- Isn't this the same thing as the security review?
- Who actually owns AI governance in a 200-person company?
- What actually changed with the EU AI Act in 2026?
- Do we need ISO 42001 certification?
- What do we actually hand an auditor on day one?
Release
- Isn't this just CI/CD with different names?
- We just edit the prompt in a dashboard. Do we really need a release process for that?
- What's the difference between this and the Operate stage in the handbook?
- How do you roll back a retrieval index that already got rebuilt?
- What counts as a severity-1 release incident if the system never actually crashes?
- Do we need all six surfaces before we ship anything?
Security review
- Is this legal advice?
- Do all four gates apply to every AI feature?
- We're pre-launch, is it too early to prep for this?
- What if we don't have a good answer to one of these yet?
- How is this different from a generic AI security checklist?
- Can a Ship Audit or Readiness & Diligence engagement help with this directly?
Cost and proof
What AI actually costs to run, and the record of what we shipped and what broke.
Learning the craft
The Applied AI Handbook and the pattern library, plus the eval stack and the failure-mode field guide.
Evals
- How big does a golden set actually need to be?
- Do we need all six layers on day one?
- Can we just use an LLM to judge our own output?
- What's the difference between an eval and observability?
- Isn't this a lot of infrastructure for a small feature?
- How is this different from the eval methodology on Capabilities?
From the field
Questions a specific insight or comparison raised, answered inside that piece.
No question matches that search. Try a different word.
Still stuck? Search the whole site, or ask us the question directly.