If you already know which jobs waste the most time, you need an implementation consultant: someone who automates specific processes and hands them over working. You need a strategy consultant when your team can't agree what matters, when AI changes what you sell or how you price it, or when a lender, investor or regulator wants a written plan first.
The labels are loose, and plenty of consultants do both, so judge by the deliverable rather than the job title. Strategy work ends in a document: priorities, risks, budget and sequence. Implementation work ends in something running in your own accounts, plus a handover. A disclosure: my own consultation is implementation-focused, so test this advice against the decision table below rather than taking my word for it.
What each one leaves you holding at the end
The quickest way to tell the two apart is to ask what exists on the last day that didn't exist on the first.
| Strategy consultant | Implementation consultant | |
|---|---|---|
| Question it answers | What should we do with AI, in what order, and what's off limits? | How do we make this particular process work with AI, reliably? |
| Typical deliverables | Ranked priorities with estimated value, rules for data and risk, budget ranges, a sequenced plan with owners | Working automations in your accounts, final prompts, test results, a runbook and a handover session |
| How you know it's done | Every priority has an owner, a number and a date, and you could brief a builder from it | Agreed tests pass on real data, and your staff can run, pause and check it |
| Who does the work afterwards | You, or an implementer you hire next | Your staff, with optional support |
| Risk if it's the wrong fit | A good document nobody acts on | A well-built automation of the wrong process |
Both risks are real, and both are expensive in different ways. A strategy nobody acts on costs its fee. An automation of the wrong process costs its fee plus the time your team spends working around it. For a fuller picture of the build side, see what an AI implementation consultant actually does.
Eight situations and which consultant fits
| Your situation | Better fit | Why |
|---|---|---|
| You can name the task that eats the most hours | Implementation | The problem is known; the value is in building and handing over |
| Partners or managers disagree about priorities | Strategy, kept short | Building before agreeing spends money on one person's preference |
| AI could change your product, service or pricing | Strategy | That's a business-model decision before it's a tooling one |
| A lender, investor or buyer wants to see an AI plan | Strategy | They want a document with reasoning, not a demo |
| Staff already use assorted AI tools and you're uneasy about data | An audit first, then either | You need to know what's happening before deciding what should |
| You want AI to help screen job applicants | Strategy, plus legal advice | Recruitment is on the EU AI Act's high-risk list, with those duties deferred to 2 December 2027; a written policy should come before any tool |
| Nobody in-house could maintain what gets built | Implementation with a strong handover, or a done-with-you approach | The skill has to stay in the business when the consultant leaves |
| You're a sole trader using two or three tools | Often neither yet | A course, a free call or an afternoon of DIY may cover it |
The fifth row deserves its own note. If you're not sure what's already running, an AI audit or a readiness assessment may be the right first purchase, before either kind of consultant. And if the done-with-you option in row seven appeals, done-for-you versus done-with-you implementation compares the two styles.
A garden centre choosing between the two, with the numbers
Take an illustrative garden centre run by two partners, with 18 staff at the spring peak, a shop, a small cafe and click-and-collect orders. Its spring problems split cleanly once they're written down.
The operational load. In April the centre receives about 640 customer emails: plant availability, delivery charges, opening hours and photos of struggling plants. At about 5 minutes each, that's 53 hours of staff time; at $18 an hour including on-costs, around $960 in April alone. Click-and-collect confirmations add another 400 messages a month in spring, copied and pasted at about 2 minutes each: 13 more hours. These are known processes with known volumes. That's implementation territory.
The disagreement. One partner wants an AI "plant doctor" chat on the website that diagnoses problems from photos. The other would rather spend the same money on a paid delivery service. Neither has data: nobody knows how many customers want diagnosis versus delivery, or what a wrong diagnosis would cost in refunds and reputation. That's a strategy question.
A sensible answer is to do both in the right order. An implementation consultant starts on the email load, and the first step of that work, sorting a month of emails into categories, produces the data the strategy question needs. Suppose the sort shows 120 of April's 640 emails were plant problems and 85 asked about delivery. Now the partners are arguing about real numbers, and a short strategy session, or just a well-prepared meeting between them, can settle it. Paying for a separate strategy engagement before the email count would have meant paying someone to guess.
The general lesson: when a strategy question depends on facts your operation already generates, collect the facts first. When it depends on judgement about the market, your appetite for risk or what you want the business to become, no amount of automation data will settle it, and strategy help is worth paying for.
Why a solo trainer needs neither kind, for now
A solo personal trainer with 30 clients, a booking app and a $20-a-month AI assistant has a different problem. The processes are small (weekly check-ins, reminders, programme write-ups), the trainer is the only decision-maker, and there's no one to disagree with. A strategy consultant would produce a document for an audience of one. An implementation consultant might be useful for a single job, such as connecting the booking app to a check-in sequence, but a free call or a few evenings with the booking app's own settings may cover most of it.
For this business the honest answer is: set up what your tools already do, then revisit in six months. If you're unsure where your own business sits, choosing between a course, a 1:1 call, an audit or a project sets out the cheaper options alongside both kinds of consultant.
When AI changes what you sell: a tattoo studio's strategy question
Row three of the table is the easiest to underestimate. Picture a two-artist tattoo studio considering AI-generated design previews: a client describes an idea, the artist produces AI variations in minutes, and the client picks one to refine by hand. Nothing here automates an existing process. It changes the product, and the questions that follow are strategic. Will clients value it, or feel short-changed next to an artist's original drawing? Who owns a design that started as AI variations, and what does the studio say when a client asks for something "in the style of" a living artist? Should consultations be priced differently?
An implementation consultant could set up the tools in an afternoon; that part is easy. Whether to offer it, how to describe it to clients and where the line sits are decisions for the owners, possibly with legal advice on the copyright side. That's the kind of question where a structured strategy conversation earns its fee even in a very small business, because a mistake shows up in the studio's reputation, not in a spreadsheet.
Strategy work that earns its fee
Good strategy work is short and specific. If a strategy consultant can't show you a sample deliverable that looks something like the excerpt below, be cautious. An illustrative priority list from a strategy engagement for the garden centre:
| Priority | Estimated value | Main risk | Owner | Decision by |
|---|---|---|---|---|
| 1. Automate email triage and click-and-collect confirmations | 60-70 staff hours in each peak month | Wrong stock information in replies | Operations partner | Start in February |
| 2. Trial paid delivery on weekends | Unknown until an 8-week trial | Van and driver costs | Retail partner | After spring |
| 3. Plant-problem guidance on the website | Depends on email volumes | Wrong diagnoses; refunds | Plant team lead | Review after the email data |
| Off limits for now | AI-generated product photos for plants (customers judge plants by real photos); any AI use in hiring seasonal staff | |||
Notice what makes it useful: it ranks, it attaches numbers or admits they're unknown, it names owners and dates, and it says what not to do. A 40-page report without those five things is an education, not a strategy. The format in writing a one-page AI strategy is a good benchmark for what should fit on a page.
Implementation work you can check line by line
Implementation is easier to judge because it either works or it doesn't, provided you agreed in advance what "works" means. For the garden centre's email project, acceptance criteria might read:
- Sorts 50 real April emails into the agreed categories with no more than 3 errors, and flags anything unclear for a person.
- Drafts replies to availability questions using only the live stock sheet; never states a price or date that isn't on it.
- Sends click-and-collect confirmations automatically within 10 minutes of an order, with the collection time taken from the order record.
- Runs in accounts the garden centre owns, with failure alerts going to the operations partner's inbox.
- Two staff members can pause it and restart it, following the runbook, without help.
Every line is testable by someone who isn't technical. If an implementation proposal can't be turned into lines like these, it isn't specific enough yet, whoever wrote it.
Questions that reveal which kind of consultant you're talking to
Titles don't tell you much, so ask questions whose answers do. Five that work:
- "What did you hand over at the end of your last engagement?" A strategy consultant describes a document; an implementation consultant describes accounts, workflows and a runbook. Someone who describes "insights" may be neither.
- "What happens in the first week?" Strategy starts with interviews and goals. Implementation starts with access, sample data and a walkthrough of the process.
- "Who maintains it afterwards?" "You'll have a plan to follow" is strategy. "Your team, using the runbook" is implementation.
- "What's the smallest engagement you'd take on?" Implementers can usually scope a single process; strategy work needs the whole business in view.
- "What would you advise me not to do?" Both kinds should have an answer. Anyone who can't name something to avoid hasn't thought about your business yet.
Asking both kinds the same questions also shows whether they respect the boundary. A strategy consultant offering to "knock up a quick prototype" for free, or an implementer keen to rewrite your business plan, may be stretching past what they do well.
Drafting your own strategy first with an AI assistant
Before paying for either, you can get surprisingly far by asking an AI assistant to challenge your thinking, as long as you feed it your figures and forbid it from inventing others. A prompt the garden centre partners might use:
Act as a sceptical adviser to a garden centre with two partners and
18 seasonal staff. Using ONLY the figures below, rank these three
options by likely value and risk, and list what data is missing
before we could decide.
Options:
1) Automate email replies and click-and-collect confirmations
2) An AI "plant doctor" chat on our website
3) A paid delivery service
Figures: [paste volumes, minutes per task, staff costs]
Rules: no invented statistics or industry benchmarks. Where a figure
is missing, write "unknown" and say how we could measure it.An illustrative extract from the reply:
Option 1 ranks first: the figures show about 66 staff hours in April on emails and confirmations, a known cost with a low-risk fix. Option 3 ranks second on potential, but demand is unknown; industry data suggests around 30% of garden centre customers would pay for delivery. Option 2 carries the highest risk, because an incorrect diagnosis could lead to refunds, and the number of plant-problem emails is unknown.
Most of that is useful, and the ranking of option 1 matches the numbers. But the "30% of customers" line is exactly what the prompt forbade: an invented benchmark with no source. Strike it and ask again: "Remove any figure not in my data." What's left is a fair first draft of the strategy conversation, and a clear list of the data you need. If the draft is enough for the partners to agree, you may not need a strategy consultant at all. If it shows the disagreement runs deeper than data, that's the sign you do.
Putting a strategy quote and a build quote on the same footing
The two are usually priced differently, which makes the totals hard to compare. Strategy work is often a fixed fee for a defined document and a set number of sessions. Implementation is often a fixed price per process, a discovery fee followed by a build price, or hourly with a cap. So compare what each fee buys against your own numbers instead. The garden centre's email project has a measurable value, roughly 66 staff hours in each peak month, so its quote can be judged against that. The strategy question has no comparable number until the email data exists, which is one more reason to run it second.
Whichever you buy, ask each consultant to split the quote into two lines: what you'll hold at the end, and what you'll pay every month afterwards, such as tool subscriptions, usage fees and any ongoing support. A strategy quote's second line should be close to zero. An implementation quote's second line is the running cost you'll live with, so it deserves as much scrutiny as the build price.
Signs you've hired the wrong kind
Mismatches happen because the job titles overlap. Three illustrative patterns, and how to rescue each:
- A strategy consultant who keeps building prototypes. Three months in, there's a demo chatbot on someone's laptop, no one owns it and it isn't connected to anything real. The rescue: stop the build work, ask for the priority list with owners and dates that you actually paid for, and hire implementation separately if needed.
- An implementation consultant automating a process that shouldn't exist. They've built a neat workflow to send customers a weekly stock email that the partners were about to stop sending anyway. The rescue: pause and have the priority conversation, even a short one, before the next build.
- A plan with no one to carry it out. The strategy is good, but nobody in the business has the time or skills to act on it. The rescue: either scope an implementation phase or appoint an internal owner; for ongoing senior direction some businesses consider a fractional chief AI officer.
The best protection is the same for both: agree the deliverable in writing before any work starts. If the consultant describes the end of the engagement as "you'll have a clearer picture", ask what document or working system that picture will be.
Further reads
- Do I Need an AI Consultant? 8 Signs It's Time to Get Help — Eight signs you need outside help at all, before choosing which kind.
- AI Consultant vs Agency vs Freelancer: Which Should You Hire? — Once you know the kind of work, choose the kind of supplier.
- AI Implementation Roadmap for Small Businesses: 5 Phases in 90 Days — A five-phase roadmap that combines both kinds of work.
- AI Consultant vs Your IT Support Company: Who Should Handle AI? — When your IT provider can do the implementation instead.
- How to Run a Paid Discovery Phase Before a Full AI Project — A paid discovery stage that often settles the strategy question.
- What Is the Cheapest Way to Get Expert AI Advice? — Low-cost options if you need neither consultant yet.
- AI Consultant or DIY: Which Suits a Small Professional Firm? — When a small professional firm can set up AI itself, when outside help earns its fee, and the split approach many firms settle on.
- Is Your AI Consultant Independent? Checking for Tool Commissions — The referral terms vendors publish, a sum of what a recommendation is worth to a consultant, six questions that get straight answers, and a disclosure clause.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: EU AI Act (Annex III high-risk categories and the Digital Omnibus deferral dates); vendor plan pages for tools named in the examples.