AI Implementation Consultant: The Vetting Checklist

June 5, 2026

Strategy consultant versus AI implementation consultant: a two-column comparison of deliverable, who operates after, and proof of work

Last updated: June 4, 2026

If you have approval to spend on AI implementation, three consultant pitches on your desk, and two of the decks say "AI strategy and implementation" without ever defining where strategy ends and implementation begins, this checklist is for you. Pick the wrong partner and twelve months out you own a slide deck and a security questionnaire instead of a deployed agent; the budget closed, the board has questions, and the operator who was supposed to run it on Monday morning never had a runbook. In this guide, you'll get the 7-item implementation-specific vetting checklist, the three killer questions that separate builders from resellers, and the proof of work to demand before any SOW is signed.

A scoped single-workflow agent reaches production in 6 to 10 weeks of build, or 8 to 12 weeks when the deployment is private; a transparent implementation proposal names that calendar, the integration target, and the operator before pricing the work. We use what we sell, including the agents that run Arkeo itself. Before you score a single proposal, scope the build; a free AI Assessment defines the one workflow worth deploying first so the proposals can be measured against a real specification, not a sales pitch.

Quick Answer
• What an implementation consultant does: Builds and integrates AI systems. Deliverable is a deployed workflow in production, not a slide deck.
• How they differ from strategy consultants: Strategy ends at a roadmap; implementation ends at a running system someone operates Monday morning.
• How to vet one: The seven items below plus three killer questions that separate builders from resellers.
• Why it matters: The wrong category buy produces a deck when the budget needed a deployment, and twelve months later there is still nothing to operate.

What does an AI implementation consultant actually do?

An AI implementation consultant builds and integrates AI systems into your operation; the deliverable is a deployed workflow agent running in production against your data, not a roadmap or a slide deck. The work spans scoping, model selection, integration with the ERP or CRM that holds the data, security review, deployment, monitoring, and the operating handoff to whoever runs the agent after launch. Strategy consultants advise, implementation consultants ship.

The category exists because the market matured. The Stanford HAI 2025 AI Index reports 78 percent of organizations used AI in 2024, up from 55 percent the year before; the bar moved from "are you using AI" to "is anything in production." The PwC AI Agent Survey of 300 senior US executives found 79 percent of US businesses are already adopting AI agents and 66 percent of adopters report measurable productivity gains. The other third has nothing to point to, and the difference between the two groups is almost always implementation, not strategy.

How is an implementation consultant different from a strategy consultant?

The cleanest way to see the distinction is to lay the two side by side on the artifacts that change hands at the end of the engagement. A strategy consultant earns the fee when the deck is approved. An implementation consultant earns it when a named operator inside your company opens the agent on a Monday morning and runs the workflow against live data. Anything else is a partial delivery.

Most reseller pitches blur this on purpose. The blur is good for revenue and bad for the buyer, because the proposal language sounds the same in both cases and the price is comparable. The difference shows up in the proof of work. Strategy firms cite frameworks and methodologies; implementation firms cite deployed systems that are still running and the named people who run them. The IBM IBV CEO Study of 2,000 CEOs across 33 countries found 54 percent of CEOs are already hiring for AI roles that did not exist a year ago, with "lack of expertise" cited as the top barrier; the implementation gap is the expertise gap, and it is not closed by a deck.

What is the 7-item vetting checklist for an AI implementation consultant?

The checklist below is implementation-specific. It deliberately ignores the items that belong on a strategy-firm scorecard (frameworks owned, partner networks, industry research) and focuses on what proves a firm can build, integrate, and hand off a working system. Run every shortlisted firm through all seven before any short list becomes a contract.

THE VETTING CHECKLIST

Seven items every implementation buyer should demand

A builder answers all seven without hedging. A reseller answers two and pivots to the demo.

ITEM 01

Deployed systems in production today

Demand named companies and the workflow currently running. Anonymized references and screenshot decks do not count. Red flag: Every case study reads "a Fortune 500 client in financial services."

ITEM 02

Integration history with your stack

Ask which exact ERP, CRM, ticketing, or document system they have integrated against, and how many times. Red flag: "We integrate with anything" with no specific system named.

ITEM 03

Operations plan post-launch

Require the runbook, the monitoring cadence, and the named operator before any build SOW is signed. Red flag: The proposal stops at "go-live" with no operating model attached.

ITEM 04

Security and data sovereignty answer

Where will the data live during training, inference, and logging, and who can see it. Get the answer in writing. Red flag: "We use a leading public model" with no on-premise or private option.

ITEM 05

Hand-off model

Managed services backstop, internal team trained to operate, or both. Decide before kickoff, not at launch. Red flag: "Your team takes over after go-live" with no training plan attached.

ITEM 06

Pricing transparency

Fixed scope, time and materials, or retainer, with a clear answer on which model the build uses and why. Red flag: Open-ended T and M with no ceiling and no defined exit criteria.

ITEM 07

Kill-criteria willingness

Ask them to write down the measurable result by which date that would cause the build to be halted on evidence. Red flag: Refusal, deflection, or "we have never needed to halt a project."

A builder answers all seven without hedging. A reseller answers two and pivots to the demo.

The pricing-transparency item is the one most buyers underweight, and it has the most leverage. In Arkeo's build experience, a scoped single-workflow agent runs about $15,000 to $40,000 and reaches production in 6 to 10 weeks, or 8 to 12 weeks when the deployment is private or enterprise-grade. Off-the-shelf copilots come in at roughly $20 to $30 per user per month and go live in days. The first quick win typically lands inside 30 to 90 days. Picture a 280-person specialty distributor scoping a quoting-assistant agent: a transparent fixed-fee proposal will name the price band, the calendar weeks, the integration target (an existing ERP), and the deliverable (a quoting workflow with a named operator). A reseller proposal at the same scope will price by user seat and quietly assume your internal team will run it.

Score real proposals against a real scope
One 60-minute free AI Assessment defines the single workflow worth deploying first, so every implementation proposal you receive is scored against a specification you wrote, not a sales pitch you absorbed.
Book Your Free AI Assessment →

Which three questions separate builders from resellers?

The seven-item checklist is the structural pass. The three questions below are the diagnostic that exposes the difference between a firm that has shipped systems and a firm that has only sold them. Ask them on a call, listen for the texture of the answer, and write down which side of the table the answer comes from.

THE KILLER QUESTIONS

Three questions and the two kinds of answer

Builders answer in specific people, dates, and broken parts. Resellers answer in decks and logos.

QUESTION 01

What is a system you deployed 12+ months ago that is still in production?

Builder answer: "Yes, here is the company and the industry, here is what they run today, here is the operating cadence and who watches it."

Reseller answer: "Let us pull up a deck of logos we have worked with."

QUESTION 02

Who operates it now?

Builder answer: "The client's named owner, with our managed-services backstop on the model-health side."

Reseller answer: "The client team handles it after launch."

QUESTION 03

What was the first thing that broke?

Builder answer: "A model update changed an output format. A permission scope was too wide. Here is what we patched."

Reseller answer: "Nothing has broken."

The questions that separate builders from resellers are the ones nobody scripts.

Most businesses think AI implementation is a vendor choice. It is closer to a partner choice, because the work continues after launch. The blunt truth is that AI agents break, regularly, and the value of the implementation consultant after go-live is greater than the value of the implementation consultant before go-live. BCG's Where's the Value in AI? report from October 2024 found 74 percent of companies struggle to capture value from AI, and the constraint is rarely the model; it is the absence of an operator who notices when the agent drifts. The Deloitte State of Generative AI Wave 4 study of 2,773 C-suite respondents found more than two-thirds expect 30 percent or fewer of their generative AI experiments to scale within three to six months. The implementation consultant who cannot answer the operations question is the one whose pilot becomes that statistic.

What proof of work should you demand?

Proof of work is what every item on the checklist eventually points back to. A builder demonstrates it on screen during the proposal call: a live system, the workflow it runs, the people who operate it, and the on-call rotation that catches it when it breaks. Anonymized references and award screenshots are not proof of work; they are marketing. Picture a regulated-services buyer evaluating two AI implementation firms for a claims-summarization workflow: one offers a 20-minute screen share of a 14-month-old claims summarization agent still in production at a named client, with the operating cadence walked through live; the other offers a logo wall and a security questionnaire. The first one is the build partner. The second one is selling a deck.

The cross-cluster posts on the general AI consultant pillar and how to hire an AI consultant cover the broader vetting work that applies across categories, and the buyer weighing whether to hire externally at all should start with the AI strategy consultant decision. This post stays narrowly on the build half of the table because most lost-budget stories trace back to a buyer who used a strategy scorecard on an implementation engagement and accepted a deck where a deployed system should have been.

What is the operations question most buyers skip?

The operations question is who runs the agent on the Monday after the project plan ends. Most buyers skip it because the proposal does, and the proposal does because the reseller does not have an answer. Arkeo deploys a private AI workforce on the client's infrastructure where data never leaves the building, then operates the system under the Assess, Deploy, Manage model so the agent has a named owner and a model-health cadence the day it goes live. We use what we sell, which is to say the same private agents that run Arkeo's own operation are the agents Arkeo deploys for clients. The wider methodology sits inside the pillar on enterprise AI strategy; this spoke deliberately stops at the build-partner vetting layer, because operations is what the vetting checklist is actually measuring.

Turn three proposals into a scored decision
The free AI Assessment names the single workflow worth deploying, the operating model that will keep it running, and the seven-item checklist applied to a real specification so every implementation proposal is scored on the same scale.
Book Your Free AI Assessment →

Frequently Asked Questions

What does an AI implementation consultant do?

An AI implementation consultant builds and integrates AI systems into a business operation. The work spans scoping the workflow, selecting the right model, integrating with the ERP or CRM that holds the data, completing the security review, deploying the system, monitoring it after launch, and handing operations to a named owner. The deliverable is a deployed agent running in production against live data, not a roadmap or a slide deck.

How is an AI implementation consultant different from a strategy consultant?

A strategy consultant advises and delivers a roadmap or a slide deck; the engagement ends when the deck is approved. An implementation consultant builds and integrates a system; the engagement ends when a named operator inside the client company runs the workflow against live data on a Monday morning. The proof of work also differs: strategy firms cite frameworks and methodologies, while implementation firms cite deployed systems still running in production and the named people who operate them.

How much does an AI implementation consultant cost?

In Arkeo's build experience, a scoped single-workflow agent runs about $15,000 to $40,000 and reaches production in 6 to 10 weeks, or 8 to 12 weeks when the deployment is private or enterprise-grade. Off-the-shelf copilots that complement the build come in at roughly $20 to $30 per user per month and go live in days. The first quick win typically lands inside 30 to 90 days. Transparent pricing names the model (fixed, time and materials, or retainer), the calendar weeks, the integration target, and the deliverable up front.

What proof of work should a buyer demand from an AI implementation consultant?

A live demonstration of a system the firm built that is still running in production 12 months later, with the named client (or, when bound by confidentiality, the verifiable industry and workflow), the operating cadence, and the person who operates it. A logo wall, an anonymized case study, and an award screenshot are marketing, not proof of work. The single best diagnostic question is: "What was the first thing that broke after go-live?" A builder answers with a specific incident; a reseller says nothing has ever broken.

Who operates the AI system after an implementation consultant finishes?

A builder names the operator at kickoff, before any code is written, and the answer is one of three patterns: the client's internal lead with a managed-services backstop on model health, a fully internal team trained by the consultant during the build, or a managed-services arrangement for an agreed term while the internal capability is built. A reseller defers the question to "after go-live," which is the same as not having an answer. The post-launch operating model is the load-bearing decision; without it the deployment stalls.

What is the difference between an AI implementation consultant and an AI integrator?

The terms overlap heavily in practice. AI integrator is usually used to describe a firm whose primary skill is wiring an existing AI tool into a client's system stack; AI implementation consultant is the wider category that also includes scoping the workflow, selecting the model, designing the operating model, and handing off to a named operator. A buyer running the seven-item vetting checklist will get the same yes-or-no signal from either label; the work is what matters, not the title on the proposal.