The question “project or retainer?” is the wrong question.
The right question is “what phase is my company in, and what does that phase require?”
Phase 1, building AI Foundations, is a defined scope with a specific deliverable: a complete context pack, workflow documentation, a configured workspace. That is a project.
Phase 3, building automated workflows and connecting them, is an evolving scope where each workflow build depends on what the previous one produced. That is a retainer.
Most companies need a project first and a retainer after.
This article defines what each engagement model produces, which phase of AI development each is appropriate for, and the four conditions that determine which model a company should start with.
What the project model produces: and where it fits
What a project engagement is
A fixed-scope engagement with a defined set of deliverables, a defined timeline, and a defined fee. The engagement ends when the deliverables are complete. The scope does not change during the engagement (or changes through a formal scope adjustment process).
The three AI phases that suit a project structure
Phase 1 AI Foundations project (3 to 6 weeks, $15,000 to $40,000)
Scope: context pack including voice guide, client archetypes, and decision rules, plus workflow specification documents for five to eight core workflows, configured shared workspace, and AI system owner briefed and onboarded.
Why it suits a project: the deliverables are known before the engagement starts. The context pack contents are predictable. The workspace configuration is a standard set of steps. The engagement executes a defined build rather than evolving as it discovers.
Phase 1 and Phase 2 project (8 to 14 weeks, $30,000 to $80,000)
Scope: all Phase 1 deliverables plus role-specific training for every AI-using team member, adoption tracking log installed, blended acceptance rate above 75% across deployed workflows, and the AI system owner running the maintenance cadence independently.
Why it suits a project: the training scope is definable (number of role types × training session length), the acceptance rate target is measurable, and the engagement end condition is specific.
The scope may extend slightly if adoption takes longer than expected, but the variability is bounded.
Specific automation project (4 to 8 weeks, $12,000 to $30,000 per workflow or workflow cluster)
Scope: one to three specific automated workflows, from mapping through specification through build through testing and proof of 80%+ acceptance rate for 30 days.
Why it suits a project: the specific workflows are defined before the engagement starts, the build process is standard, and the acceptance rate target is measurable.
This structure is appropriate for a company that has completed Phase 1 and Phase 2 and wants to build specific Phase 3 automations without committing to a full Phase 3 retainer.
The project model’s limitation
The project model cannot accommodate scope evolution well.
If the Phase 3 build reveals that two additional context pack entries are needed to make the automation work, this is either a scope addition (which costs more) or something left for the company to handle after the project ends.
In a retainer, the same discovery is addressed in the next weekly session without a formal scope process.
What the retainer model produces: and where it fits
What a retainer engagement is
An ongoing monthly engagement where the consulting firm is available for a defined level of work per month: build work, maintenance oversight, improvement cycles, and strategic guidance, at a fixed monthly rate.
The scope evolves based on what the company’s AI system requires each month.
The two AI phases that suit a retainer structure
Phase 3 retainer (months 4 to 8 and beyond, $8,000 to $12,000/month)
Monthly scope: building automated workflows from the Phase 2 proven manual workflows, knowledge base population, context pack maintenance and updates, adoption tracking review and improvement cycles, AI system owner support.
Why it suits a retainer: Phase 3 build work is inherently iterative. Each automation build informs the next one. Discoveries in one workflow affect the specification of another.
The scope expands naturally as the company’s AI system matures. A project structure that tries to scope twelve months of Phase 3 work in advance will encounter scope changes at every turn. A retainer accommodates this evolution naturally.
Phase 4 retainer (months 8 to 18, stepping down to $3,000 to $6,000/month)
Monthly scope: connecting Phase 3 automations into agent chains, supporting the AI system owner in running the connected system, responding to escalations, and quarterly strategic reviews.
Why it suits a retainer: Phase 4 work is ongoing oversight and incremental development, not a defined build project. The consulting firm’s role is advisory and responsive rather than constructive. The retainer rate steps down to reflect this reduced intensity.
The retainer model’s risk
The retainer model’s primary risk is drift, activity without measurable output.
Unlike a project, which ends when deliverables are complete, a retainer continues when the specific deliverables for the month are unclear.
A retainer without a monthly deliverable commitment is an ongoing fee for undefined access, which can produce real value or can produce the “meetings replacing milestones” drift described in this series.
The retainer that lacks a monthly deliverable agreement is not a consulting engagement — it is a subscription to availability.
The mitigation: a monthly deliverable agreement built into the retainer structure.
At the start of each month, the consulting firm and the company agree on the specific outputs for that month: two new workflow specifications, three automation builds, a context pack revision addressing the previous month’s adoption log patterns.
The monthly deliverable agreement converts the retainer from a fee for access into a fee for specific production.
The four conditions that determine which model to start with
Condition 1: AI maturity level
Starting from zero (no context pack, no documented workflows, no team training): start with a project. The Phase 1 and Phase 2 scope is defined and bounded. A retainer starting at zero would be a retainer for foundational work that should be scoped as a project.
Starting from Phase 2 completion (context pack in place, team trained, workflows at 75%+ acceptance rate): a retainer for Phase 3 work is appropriate. The Phase 3 scope evolves based on what Phase 2 produced. A project structure would require re-scoping every four to six weeks.
Starting from somewhere in between (partial foundation, some training, inconsistent adoption): a project to complete and stabilise the foundation, a targeted Phase 1 remediation and Phase 2 completion project, before moving to a Phase 3 retainer.
Condition 2: Scope clarity
Can the specific deliverables for the next 8 to 14 weeks be defined clearly?
If yes: a project is appropriate. If the scope is clear enough to write a deliverables list, a project structure provides better budget certainty and clearer accountability.
If no: the scope evolves too quickly for a project. Phase 3 automation builds often reveal that the context pack needs updating mid-build. A retainer accommodates this. A project creates scope disputes.
Condition 3: Internal capacity to maintain momentum between projects
If the company uses a project model for Phase 1 and Phase 2, does the internal AI system owner have the capacity to maintain the foundation and progress toward Phase 3 readiness while the next project is scoped and engaged?
If yes: the project model works. The gap between projects is managed internally.
If no: the gap between project engagements produces context decay, adoption decline, and the need to re-establish what was built before the next project can build on it. In this case, a retainer that provides continuity between phases is more efficient than a series of projects with gaps.
Condition 4: Budget preference vs continuity preference
Budget certainty preference: companies that need to know exactly what the AI consulting investment will cost in a given quarter will prefer the project model. The fee is known before the engagement starts and does not change.
Continuity preference: companies in active Phase 3 or Phase 4 development who need the consulting firm consistently present will prefer the retainer. The continuity of engagement is worth the open-ended commitment.
Most companies prefer budget certainty for Phase 1 and Phase 2, and continuity for Phase 3 and beyond. This naturally produces the hybrid model.
The hybrid model: project for Phase 1 and 2, retainer for Phase 3 and beyond
Why the hybrid model is optimal for most companies
The hybrid model aligns engagement structure with AI development phase.
Phase 1 and Phase 2 are defined and bounded: project appropriate. Phase 3 and Phase 4 are evolutionary and ongoing: retainer appropriate.
Using a project for Phase 1 and Phase 2 and a retainer for Phase 3 and beyond gets the benefits of each model at the right phase.
What the transition from project to retainer should look like
The transition requires a specific decision, usually at the Phase 2 completion review, that the company wants to continue into Phase 3 with the same firm. The transition meeting covers four things.
- Phase 2 completion verification: are the acceptance rate targets met? Is the AI system owner independent?
- Phase 3 scope discussion: what are the first three automation builds for Phase 3?
- Retainer rate and monthly deliverable agreement: what does the firm deliver each month, and at what rate?
- The first monthly deliverable agreement for Phase 3, written and agreed before the retainer starts
The retainer rate at Phase 3 start
The transition from a Phase 1 and Phase 2 project ($30,000 to $80,000 total) to a Phase 3 retainer ($8,000 to $12,000/month) represents a significant change in fee structure.
Evaluate this transition at Phase 2 completion using two criteria.
Criterion 1: did the Phase 1 and Phase 2 project produce what it was supposed to? If not, resolve the deficit before entering a retainer.
Criterion 2: is Phase 3 automation work the right next investment for this company right now? Some companies are better served by stabilising Phase 2 adoption internally before moving to Phase 3.
When not to move to a retainer
The company that completed Phase 1 and Phase 2 and is now at 75%+ blended acceptance rate has a stable, self-maintaining foundation.
If the AI system owner is capable and has the capacity, Phase 3 automation builds can be scoped as individual projects rather than as a retainer.
This gives the company budget certainty on each build while maintaining the option to pause between builds.
The retainer is right when Phase 3 and Phase 4 work is continuous enough that project scoping between each build would create friction without providing meaningful benefit.
For many mid-market companies with 10 to 20 targeted automations to build, a series of small projects is equally effective and provides more budget control.
Common questions on project vs retainer
”Can the retainer be paused between months without penalty?”
Depends on the contract. A well-structured retainer contract should include a pause clause, typically a 30-day notice period to pause, with the option to resume at the same rate within 90 days.
The pause clause is important because Phase 3 build cycles sometimes need to be interrupted by business changes.
A contract without a pause clause creates pressure to continue the retainer even when the company is not ready for the next build cycle.
”What if the Phase 1 and Phase 2 project takes longer than expected: does this affect the transition to retainer?”
The transition should happen when Phase 2 is complete to the required standard, not when the scheduled end date arrives. If Phase 2 takes 14 weeks instead of 10, the retainer starts at week 14, not week 10.
Entering the Phase 3 retainer before Phase 2 is complete means paying Phase 3 retainer rates for Phase 2 completion work.
The transition trigger should be a system state, not a calendar date.
Build the Phase 2 completion condition into the transition trigger.
”What should the monthly retainer contract include to prevent drift?”
Three specific elements:
- Monthly deliverable agreement: specific outputs agreed at the start of each month, signed off by both parties
- Monthly review meeting: 30 to 45 minutes at the end of each month to assess whether the agreed outputs were produced and to agree the next month’s outputs
- Acceptance rate floor: if any deployed workflow falls below 70% acceptance rate and remains there for two consecutive monthly reviews, the retainer must include remediation of that workflow before adding new automation builds
”What is the minimum retainer commitment before I can evaluate whether to continue?”
Three months. The first month is transition and setup. The second month is the first complete build cycle. The third month is the first evaluation of whether the build cycle is producing the expected output quality.
Evaluating a retainer after one month is evaluating the transition, not the engagement. Three months provides enough data to assess whether the monthly deliverable agreement is being met and whether the acceptance rates are improving as expected.
Want to see the specific project scope and retainer structure that fits where your company is right now?
Project or retainer is a phase decision, not a preference decision.
Phase 1 and Phase 2 are project-appropriate: defined scope, specific deliverables, bounded timeline and fee. Phase 3 and Phase 4 are retainer-appropriate: evolving scope, ongoing build cycles, continuous presence.
The company that chooses correctly gets the budget certainty it needs for the foundation and the continuity it needs for the build. The one that puts Phase 1 on a retainer is paying for flexibility it does not need. The one that puts Phase 3 on a project is creating scope disputes for work that is inherently iterative.
Path one: map your current phase to the right model. Assess the four conditions in this article against your current situation. If scope clarity is high and AI maturity is low, start with a project. If scope clarity is low and AI maturity is Phase 2 complete, move to a retainer.
Path two: bring in a partner. Phos AI Labs structures Phase 1 and Phase 2 as a fixed-scope project and Phase 3 and beyond as a monthly retainer with a monthly deliverable agreement built in. The right structure for your current phase is part of the first conversation. We have run 400+ AI engagements. Clients include Zapier, Coca-Cola, Medtronic, Dataiku, and American Express. Thirty minutes, no deck. Start here.