Question 1

Strategize: A 50-person professional services firm wondering where to start.

Accepted Answer

The firm has watched competitors talk about AI for two years and feels behind. Leadership is not
sure whether to hire an AI lead, buy off-the-shelf tools, or wait. The engagement runs four to six
weeks: structured interviews with operational leads, a readiness assessment, and a written
deliverable identifying the two or three highest-leverage opportunities, a build/buy/partner
recommendation for each, and a phased plan. By the end, the firm has a defensible position to take
to its board, including, often, a recommendation not to act on certain opportunities yet.

Question 2

Implement: A regional manufacturer rolling out an internal knowledge base.

Accepted Answer

The company's technical service team fields hundreds of repeat questions a month, most of which
are answered somewhere in 20 years of accumulated PDFs, emails, and shop-floor notes. The
engagement covers source document curation, retrieval-augmented generation pipeline design,
evaluation against a held-out set of real questions, integration with the team's existing ticketing tool,
and two training sessions for the people who will own it day to day. Six to ten weeks. Deliverable: a
working pipeline, documentation, and a team trained to maintain it. A 30-day stabilization period
follows deployment.

Question 3

Build: A specialty inspection company whose vision problem doesn't fit off-the-shelf models.

Accepted Answer

The company inspects a narrow class of mechanical parts where defects are rare, geometrically
subtle, and not represented in any public dataset. Standard vision APIs miss them. This is the case
where Build is genuinely warranted. The engagement covers dataset construction, custom model
development, validation against the company's existing inspection criteria, and production handoff
with full documentation. Most prospective clients are talked out of Build engagements during
scoping. Build is reserved for problems where off-the-shelf options have been tested and found
insufficient. A 30-day stabilization period follows deployment.

Question 4

Sustain: An organization 18 months into an AI deployment, unsure if it's still doing its job.

Accepted Answer

A small operations team deployed an internal AI assistant a year and a half ago. Initial results were
strong, but no one has audited it since. Users have quietly stopped using parts of it. Others rely on
outputs that may have drifted. The Sustain engagement covers a structured evaluation against
original success metrics, identification of drift or degraded performance, recommended fixes, and a
maintenance cadence that does not require a full-time hire. Sustain often runs as a recurring
quarterly check-in rather than a one-time project.

Engagement Timeline.

Additional Info

Strategize: A 50-person professional services firm wondering where to start.

Implement: A regional manufacturer rolling out an internal knowledge base.

Build: A specialty inspection company whose vision problem doesn't fit off-the-shelf models.

Sustain: An organization 18 months into an AI deployment, unsure if it's still doing its job.