Question 1

How is this different from hiring a consulting firm?

Accepted Answer

Three differences that matter in the MSA, not the pitch deck. First, the engineer you meet in the scoping call is the engineer writing the PRs, and their name is in the statement of work. Second, there is no platform-license attachment, we pick your stack and sign away IP in writing. Third, the first PR lands in 7 days or billing pauses, and the prototype has a pass-fail rubric at week 2 with a refund-and-exit gate. Partner-plus-grads firms structurally cannot do any of those three.

Question 2

How do you handle the MIT NANDA "95% of AI pilots fail" concern?

Accepted Answer

We quote the rubric we miss on. Week 2 prototype on your eval, or refund. Week 6 production or pause billing. We cite Monetizy.ai (1 week to 8K/day), Upstate (400K+ emails), and OpenLaw (publicly released) because those are the shape of outcome we commit to, measured in shipped volume and not in deliverable pages. We also turn down scopes we do not think we can land.

Question 3

Who owns the IP and the repo?

Accepted Answer

You do, from clause one. No proprietary agent framework, no platform license, no vendor-attached runtime, no data lock-in. Our contracts are explicit: your repo, your weights layer, your eval harness, your runbook, yours in writing before the engineer starts week 1.

Question 4

Do you lock us into a specific model or cloud vendor?

Accepted Answer

No. We pick model (Anthropic, OpenAI, Bedrock, Vertex, Azure OpenAI, open-weight) and orchestration (LangGraph, Pydantic AI, custom) per problem. The agent we ship runs where your security review already cleared. If Anthropic's per-token pricing changes mid-engagement, we document the fallback model and cost envelope; the FinOps deliverable is in scope.

Question 5

What does an engagement cost?

Accepted Answer

Fixed weekly rate, published up front, scoped to a specific outcome. Most engagements run 4-8 weeks and price like hiring one staff-level MLE plus a tech lead at-cost, without the 58-day hiring cycle, the $400K permanent liability, or the ramp. You will get a fixed fee in writing at the end of the free scoping call.

Question 6

How do you address EU AI Act obligations?

Accepted Answer

EU AI Act high-risk obligations go fully live August 2, 2026. The scoping call asks where your use case falls on the risk classification. If it is high-risk, we build the documentation set (risk management system, data governance, logging, human-oversight design) into the delivery and the week-6 handoff includes the compliance artifact that would ship to a notified body. If it is minimum-risk, we still build the audit trail so you are not stuck if classification changes.

Question 7

Do you bring a proprietary agent framework?

Accepted Answer

No. We write boring agent code on open-source primitives your team already knows. We would rather hand off 800 lines of Pydantic and LangGraph you can read on day one than 8,000 lines of a custom DSL that dies when we leave.

Question 8

What about MCP and A2A?

Accepted Answer

MCP-native, A2A-compatible. MCP now has 10,000+ servers and 97M monthly SDK downloads. A2A v1.0 includes Microsoft, AWS, Salesforce, SAP, ServiceNow. If you buy an agent that does not speak these protocols, you will be re-platforming inside 18 months. Our default build is protocol-native so the agent fits your existing integrations without a wrapper layer.

Question 9

What happens after you leave?

Accepted Answer

Three things land in your repo at handoff: the runbook (what to do when it breaks), the eval harness (how to tell if a change made it worse), and the architecture doc (why it was built this way). The engineer who ran the engagement is available for paid 2-hour consults at a published capped rate for 12 months after departure. No retainer, no surprise bills.

Question 10

Can we talk to a past client?

Accepted Answer

Yes, after the scoping call. We make the introduction to someone working in a shape similar to yours. It is a 20-minute call with the engineer who owned the rollout, not a marketing panel.

Question 11

Where are the engineers based?

Accepted Answer

Remote-first, overlap with your working hours. We publish the engineer's timezone and daily working window in the statement of work. Standup is async in your Slack, with a live window for pairing if you want it.

Question 12

What if our scope is too small (or too big)?

Accepted Answer

Too small is fine, small-scoped production agents are our favorite engagement. Too big gets decomposed: we will scope the first outcome we can land in 6 weeks and defer the rest until that is running in production. We will not accept a 6-month pilot, because the 6-month-pilot delivery model is the failure mode we are built against.

Answers for the procurement committee, not the marketing page.