The 14-day AI sprint: a better model than transformation theatre

What does a six-month "AI transformation" engagement actually produce?

Slides. A lot of them.

A typical engagement runs three phases: discovery (six to eight weeks), strategy (six to eight weeks), and implementation (three to four months). The first two phases produce documents. The third phase, in the cases the engagement makes it that far, produces a pilot. The pilot ships at month five, runs for sixty days, generates inconclusive metrics because the sample size is too small, and the engagement closes with a "phase two" recommendation that never starts.

This isn't a slight against the operators who buy these engagements. The format exists because procurement at large enterprises is built around it. Six-figure engagement, named partner, tenured staffing, executive readouts at quarterly intervals. The shape of the engagement matches the shape of the budget cycle.

The shape doesn't match the work.

So why does the longer cadence keep getting bought?

Three reasons, none of them about the work itself.

Optionality. A six-month engagement gives both sides time to learn what should actually ship. Both sides can change their minds at month two without consequence. The tradeoff — most learning happens by week three, but the engagement structure rewards staying in discovery for eight more weeks anyway.

Risk distribution. A six-month engagement spreads accountability across four executives, two committees, and a vendor. If the project fails, no single party owns the failure. Failure becomes diffuse. "We tried" is a shareable narrative; "we shipped a thing in fourteen days and it didn't work" is not.

Procurement velocity. Large-firm procurement physically cannot move faster than its quarterly review cycle. A two-week engagement doesn't fit a budget cycle that runs in 90-day chunks. Sixty-percent of why six-month engagements exist is that the procurement system can't approve anything shorter without a workaround.

These are real constraints. The fourteen-day sprint doesn't dissolve them. It moves around them — by routing the budget approval differently and accepting that the engagement structure can't fit a procurement system designed for vendor installs.

What changes at fourteen days?

Everything compresses, but not equally. Three things shrink to the cadence; three things stay structurally the same.

Compresses: the discovery loop. Day 1 to day 5 is the diagnostic. The deliverable is a 14-page dossier with named candidate workflows, a scored rubric, and a sequenced recommendation. By day 5, the operator and the build team agree on what ships and what doesn't. There is no "phase two". The phase is the sprint.

Compresses: the strategy artifact. A six-month engagement produces a 60-slide strategy deck. A 14-day sprint produces a one-paragraph thesis statement, embedded in the dossier, that names the workflow, the user, the failure mode, and the success criteria. If the thesis doesn't fit in a paragraph, the workflow is too broad and gets re-scoped before day 5.

Compresses: the staffing model. A six-month engagement runs eight to twelve people, including two executive sponsors. A 14-day sprint runs one operator, one build lead, and one named subject-matter expert from the client side. The smaller team is what makes the cadence possible — coordination overhead at twelve people exceeds the time available in a fourteen-day window.

Stays the same: the technical rigor. Schema deployments still get validated. JSON-LD still gets tested. Agent guardrails still get hard-capped. The build is not less rigorous because the cadence is shorter. The rigor is what fits in a sprint; the meeting count doesn't fit, so the meeting count gets cut.

Stays the same: the operator commitment. The operator still has to commit to three checkpoints — kickoff, day 7 review, day 14 walkthrough. A sprint that runs through a delegate fails at the first ambiguity. The cadence shrinks the timeline, not the operator's seat.

Stays the same: the failure rate. Sprints fail. Some workflows turn out to be the wrong workflow. Some data substrates turn out to be unfixable in the timeline. The honest read is that maybe one in seven sprints ships something the operator chooses to roll back. The recovery is fast — a fourteen-day commitment is recoverable in a way a six-month commitment is not.

How is fourteen days even enough?

Ask it differently — what makes you think six months is necessary?

The honest answer most consulting principals give in private is that six months isn't necessary, it's tolerable. A six-month engagement absorbs all the time available before someone has to ship something. A fourteen-day engagement forces the question "what would we ship if we couldn't keep planning?" on day 4 instead of month 4.

The full breakdown of what fits in fourteen days lives at what is a 14-day AI sprint. The condensed version — Day 1 to 5 is diagnostic, Day 6 to 14 is build-and-ship. The build day count assumes the substrate cleared the readiness check on day 5; if it didn't, the sprint cancels and the engagement reverts to substrate-fix work. Operators who try to push through unready substrate on a 14-day cadence get the same result an unfit runner gets attempting a marathon.

So when does the long cadence still make sense?

Three cases where six months is the right shape.

Multi-system migrations. Replacing the underlying CRM, document management system, or ERP can take six months because the data migration alone runs that long. AI work layered on top of the migration can fit in a sprint, but the migration itself can't.

Regulated deployments. A workflow under EU AI Act Annex III high-risk classification, or an India FREE-AI cluster classification, requires a documented impact assessment, a model risk evaluation, and a compliance review. Each of those takes weeks. The sprint can ship the workflow; the regulated deployment that wraps it takes longer.

Multi-stakeholder integrations. A workflow that touches sales, legal, ops, and finance requires sign-off from each. The sign-off cadence is a 90-day cycle in most enterprises. The sprint can build the workflow; the stakeholder alignment takes a separate, longer engagement.

In every other case — and most operator-led AI work falls outside these three — fourteen days is the cadence that fits the work.

What does the operator do with this read?

Three concrete moves.

Move 1 — calibrate against your own past engagements. If you ran a six-month AI engagement last year, look at the calendar of artifacts. When did the first thing actually ship? When did the second thing ship? Most operators find the first artifact that wasn't a slide didn't land until month three or four. The interval between "engagement starts" and "first artifact in production" is the number to compare against fourteen days.

Move 2 — re-scope the next engagement to a single workflow. Operators consistently scope AI engagements as "AI for the sales team" or "AI across customer support" rather than "AI that drafts the first-pass discovery brief from inbound contact form submissions." The narrower the scope, the more it fits a sprint. The narrower the scope, the higher the readiness score on the post-PMF checklist.

Move 3 — pick the next sprint, not the next strategy. The sequence is sprint → measure → next sprint, with the strategy emerging from the shipped work. The strategy that emerges this way is grounded in what actually shipped. The strategy that gets written before anything ships is grounded in what the consultancy thought was likely.

Where Doxia Axis sits in this

The 14-day sprint is the shape of the engagement, not a marketing position. The agency runs one operator with a build cadence that physically can't accommodate six-month engagements. The structural constraint became the methodology.

Tier 0 is the five-business-day diagnostic. Tier 2 is the 14-day sprint. Tier 3 is the retainer that runs three sprints over 90 days, with measurement windows between them. The retainer is not a six-month engagement compressed — it's a sequence of fourteen-day sprints with explicit measurement and re-scoping windows. The cadence stays.

Where to go from here

• Want the methodology? What is a 14-day AI sprint.
• Want the readiness check? AI readiness checklist for operators past PMF.
• Want to see it applied? Estate planning · Charlotte — 21-day engagement.
• Or just request the audit: /audit. Five business days. The diagnostic that decides whether a sprint is even the right move.