eazyware
Playbook·April 22, 2025·9 min read

How we structure AI engagements (and why)

Pilot, Build, Scale, Partnership. The four engagement tiers we offer, when each is right, and what we learned designing them.

KR
Kushal R.
Engineering lead

How we structure engagements matters more than the hourly rate. A $250/hour specialist on a well-structured engagement produces better outcomes than a $100/hour dev shop on a poorly-structured one. This post explains the four engagement models Eazyware offers — Pilot, Build, Scale, Partnership — why each exists, and how we decide which to propose for a given client.

Four tiers
Four engagement tiers Partnership Scale Build Pilot 12+ mo · $800K–3M/yr 6–12 mo · $400K–1.5M 3–6 mo · $150–600K 4–6 wk · $30–80K
Tiers stack as commitment deepens: Pilot proves feasibility; Build delivers production capability; Scale operates and extends; Partnership becomes an embedded AI function.

The four engagement models

Pilot (4-6 weeks, fixed-price)

A bounded engagement to prove feasibility or deliver a specific proof-of-concept. Fixed scope, fixed price (typically $25K-$80K), fixed timeline. Clear deliverable: a working prototype of a specific AI capability against your data. Not production-ready — the point is to learn quickly whether the approach works.

When pilots are right: you're not sure if AI can solve your problem; you need to validate with stakeholders before committing; the problem has enough ambiguity that 'let's try' beats 'let's plan.' Most clients start here.

Build (3-6 months, milestone-based)

A full production system built end-to-end. Typically $150K-$500K. Includes: eval infrastructure, observability, deployment, integration, documentation, handoff. Delivered in milestones every 2-3 weeks, with acceptance criteria per milestone.

When builds are right: you've validated the approach (often via a pilot), you're committing to production deployment, you need a complete system, not just a proof-of-concept. Most of our work is here.

Scale (ongoing, retainer + project-based)

After a Build engagement, many clients want continued Eazyware involvement. Scale engagements are retainer (fixed monthly fee for ongoing engineering capacity) plus project-based (specific new capabilities funded separately). Typically $30K-$100K/month retainer plus ad-hoc projects.

When scale is right: the AI system is core to your product or operations, you need reliable velocity without hiring full-time, you want continuity of the team that built the system.

Partnership (multi-year, embedded)

Long-term, deeply embedded engagement with one or two senior engineers fully dedicated to the client's AI roadmap. Effectively a 'hire us without hiring us.' Starts at $250K/year per embedded engineer.

When partnerships are right: AI is strategic, you don't want to build an in-house team (or can't hire the talent), you want consistent quality from a team that has worked together before. Rare but valuable when it fits.

How we pick which model to propose

Three questions drive the choice:

1. How certain are you about the approach?

High uncertainty → Pilot. Medium → Build. Low (you've done it before in another context) → Build or Scale. Skipping Pilot when uncertainty is high is the top-cited reason clients overspend in year one.

2. How much production capability do you need?

Prototype only → Pilot. First production system → Build. Continued enhancement → Scale. Deeply embedded strategic work → Partnership.

3. What's your in-house capacity?

Strong in-house team that can take over after Build → no Scale engagement needed. Small team that can maintain but not expand → Scale. No team, want us to operate the AI layer → Partnership.

What we won't do

Time-and-materials open-ended engagements. These structures incentivize us badly and produce worse outcomes for clients. Every Eazyware engagement has a defined scope, defined price, and defined deliverables. If the scope needs to change mid-engagement — which happens — we do a change order with an explicit new scope and price.

Pay-per-hour with no deliverable definition. If the client can't describe what success looks like, we haven't worked together on scoping enough. We'd rather invest a week of scoping and deliver a tight Pilot than paper over the ambiguity with hourly billing.

Payment terms

  • Pilots: 50% at kickoff, 50% at delivery. Fully refundable first two weeks if fit isn't right.
  • Builds: 25% kickoff, 50% across milestones, 25% at final acceptance.
  • Scale: Monthly, net 30, no long-term commitment beyond 3 months.
  • Partnership: Quarterly in advance, 12-month initial commitment.

How to start

Every client relationship starts with a 30-minute discovery call. No pitch. We ask about your problem, your readiness (see the readiness audit), your timeline. We recommend a model at the end of the call. If we're not the right fit, we say so and suggest alternatives. Most clients end up in Pilot or Build; a minority skip straight to Scale.

Closing

The engagement model is a decision that shapes the next six months of your AI work more than almost any other. Pick wrong and even a strong vendor will produce mediocre outcomes. Pick right and even a mid-tier vendor will produce acceptable ones. Don't skip this conversation. Our goal with every engagement is to leave the client better positioned than they started — whether that means continuing with us or running independently from here.

Read next
The 2026 guide to picking an AI vendor
Read next
The AI readiness audit: 10 questions before you write a single prompt
Read next
Build vs buy: when custom AI beats off-the-shelf
Tags
engagementpricingscopingconsulting
/ Next step

Want to talk about this?

We love debating this stuff. 30-minute call, no pitch, just engineering conversation.

~4h
avg response
Q2 '26
next slot
100%
NDA on request