Pick by what's missing. If your team lacks skills, take a course. If you need decisions about your own business, book a 1:1 call. If you don't know what's happening across your processes, tools and data, commission an audit. If you already know what to build, buy a project. Plenty of businesses need two, one after the other.
The expensive mistakes run both ways. Buying the heaviest option first, a project before you know what to build, pays someone to automate the wrong thing. Buying the lightest when the gap is capability, a single call for a team that can't yet use the tools, produces a plan nobody can carry out. The four kinds of help aren't a ladder to climb; they fix four different problems.
Four kinds of help, compared on what you walk away with
| Format | You leave with | Suits | Your time | It can't |
|---|---|---|---|---|
| Course | Skills in the people who took it | A team using AI inconsistently, unsafely or not at all | Several hours per person, plus practice | Tell you which of your processes to change |
| 1:1 call | Decisions and a written plan for your situation | An owner with candidate jobs, a stalled attempt or a quote to sanity-check | Preparation plus the call | Build anything, or inspect every system in depth |
| Audit | A written inventory of tools, automations, data use and risks, with priorities | A business with sprawling subscriptions, unowned automations or a big purchase ahead | Access, interviews, answering questions | Change anything by itself |
| Project | One job built, tested and handed over | A defined job with a clear result and an owner on staff | Briefing, supplying examples, testing, sign-off | Decide what's worth building, unless discovery is included |
Read the last column first. Each format has a job it simply doesn't do, and most disappointment with AI help comes from buying a format for the job in that column: expecting a course to produce a plan, a call to produce a working automation, an audit to fix anything, or a project to tell you whether it was the right project.
A course: when the gap is skills
A course is the right first purchase when the problem is people rather than processes. The signs: staff who've never used an AI assistant, staff using it but getting poor results, staff pasting customer details into personal accounts because nobody has shown them the company's approved tool, or a team that's nervous about it and needs time to try things safely.
Start with the free material, because much of it is good. OpenAI Academy is free to join, with courses on applying AI at work and a section for small businesses. Anthropic's Academy lists free courses such as Claude 101 and an AI fluency course aimed at small businesses. Microsoft Learn has free learning paths for Copilot, including one on preparing an organisation for Microsoft 365 Copilot that your IT provider may want to read. Paid courses are worth it when they're built around exercises on your own kind of work, or when someone will answer your team's questions as they go.
One farm shop with six staff across the counter, the office and the veg box scheme, an illustrative case, took the free route. Each person spent about three hours on a vendor course, then an hour a week for a month using the shop's business AI account on real tasks: product descriptions for the website, replies to box customers, the weekly specials board. The owner wrote a one-page "how we use AI here" note covering what never goes into the tool (card details, customer addresses) and what always gets checked by a person (anything about allergens or prices). Total cost: about 42 staff hours and no fees.
What the course didn't do was tell the shop which process to automate, and it wasn't meant to. Choosing an AI training provider for your team covers how to judge a paid course if the free ones don't fit.
A 1:1 call: when you need decisions about your business
A call is for turning a vague intention into a specific next step. The signs you need one: several ideas and no obvious first, uncertainty about whether your current tools can already do the job, a previous attempt that stalled, or a quote you'd like checked by someone who isn't selling it.
A good call ends with something written: the job to tackle first and why, the tool it'll run on (often one you already have), who owns it, and what "working" will mean. Imagine a wine merchant arriving at a call with three ideas: AI-written shelf notes, automated reorders for its trade customers, and a website chat assistant. The call ranks them. Shelf notes come first and need no consultant at all: a prompt, the producers' technical sheets and a checking step the staff can run themselves. Trade reorders come second, as a small project once the reorder rules are written down. The chat assistant is parked, because most website questions are about opening hours and delivery, which a clearer web page would answer.
That third decision is typical of what a call is for: deciding not to buy something. What happens in a 1:1 AI implementation consultation describes the call itself in more detail, including how to prepare so the time goes on decisions rather than background.
An audit: when you don't know what you've got
An audit is for businesses where AI has already spread faster than anyone has tracked. The signs: subscriptions on several people's cards, automations built by someone who has since left, staff using free chatbots with customer data, a plan to buy AI licences for everyone, or a near-miss with data that nobody can fully explain.
A useful audit covers five things and writes each one down:
- Tools and subscriptions: every AI tool in use, who pays, who uses it, and what it costs a year.
- Automations: every Zap, scenario or flow, what it does, whose account it runs in, and who would notice if it stopped.
- Data use: what business and customer information goes into which tools, and on which plans and settings.
- Access and sharing: who can see what, which matters a great deal before switching on assistants that search company files.
- Priorities: the risks to fix now and the quick wins worth taking, in order.
Consider a 20-person craft brewery, again illustrative, that commissions an audit before buying AI licences for the whole team. The findings: five overlapping AI subscriptions, two of them on personal cards and expensed monthly; two automations built by a former sales manager still running under her old login; the taproom team pasting booking lists into a free chatbot to write reminder messages; and a shared drive folder open to everyone that contains payroll exports. None of it is dramatic on its own. Together it changes the plan: cancel three subscriptions, move the automations into a company account, fix the folder permissions, and only then pick one business AI plan for everyone. What an AI audit covers and costs goes further, and AI audit versus AI readiness assessment explains which one fits a business that has barely started.
A project: when the job is already defined
A project is the right purchase when you can describe the result in a paragraph, name the systems involved, say how you'll test it, and name the person who'll own it afterwards. If any of those four is missing, a project will either stall or quietly turn into the discovery work you didn't buy.
A specialty coffee roaster with a clear brief shows what ready looks like. Its brief: "Wholesale orders submitted on our trade form are added to the roasting schedule sheet with the right roast date, the customer gets a confirmation, and orders over 20kg are flagged to the head roaster. Tested on 30 past orders. The production manager owns it." That brief can be priced as a fixed project by several suppliers and compared like for like. How to scope an AI project shows how to write deliverables and acceptance criteria when your own brief isn't that crisp yet.
Projects are also where a hidden fifth option often sits: the paid discovery phase, a short piece of work that produces the brief before anyone quotes the build. If you're somewhere between "we should automate orders" and a brief like the roaster's, that's usually the cheaper route in.
Five questions to find your starting point
Answer these in order and stop at the first one that points somewhere:
- Are staff already using AI tools you haven't approved or can't see? If yes, start with an audit, even a light one you run yourself, because every later decision depends on knowing what's already there.
- Could your team use a business AI assistant well if you gave them one tomorrow? If not, a course comes before anything else, or alongside it.
- Can you name the single job you'd most like AI to take on? If not, book a call.
- Can you write that job's result in a paragraph, with the systems named and a test for "done"? If not, a call or a paid discovery phase comes next.
- Is someone on staff ready to own it after launch? If yes, buy the project. If not, sort out the owner first, or buy the project with a support arrangement and a plan to hand it over.
Try it on an illustrative butcher with three shops. Question 1: no unapproved tools, the owner checked. Question 2: most staff already use the company's AI assistant for customer emails, so no course. Question 3: yes, the job is click-and-collect order entry. Question 4: yes, the manager can write it in a paragraph. Question 5: the shop manager will own it. Result: go straight to a project, and skip the other three formats entirely.
One question to ask before buying each format
Whichever format the five questions point to, one question to the seller tells you whether you'll get what that format is supposed to deliver:
- Course: "What will my staff do in week two, after the course ends?" A good answer includes practice on their own tasks, a way to ask questions, or a follow-up exercise. A course that ends at the certificate is often forgotten within a month.
- 1:1 call: "What will I have in writing afterwards?" You want at least the first job, the tool, the owner and how you'll judge it. If the answer is "you can take notes", take very good notes, or book with someone else.
- Audit: "What exactly will the report list, and will it rank the fixes?" An inventory without priorities leaves you to do the hardest part yourself. Ask to see a redacted example.
- Project: "What does done mean, and how will we test it?" The answer should mention your real examples, a number of cases, and what happens to anything the system can't handle.
If the seller can't answer their question clearly, that's a sign you're being sold a different format under the label you asked for, such as a sales call described as a consultation or a template described as a project.
Sequences that work, and ones that waste money
When you need more than one format, the order matters. These sequences tend to work:
- Call, then project. The call produces the brief; the project builds it. The most common route for a first automation.
- Audit, then call, then project. For businesses where AI has already sprawled. The audit tells the call what it's working with.
- Course, then call. A team that has spent a month using the tools brings much better questions to a call.
- Project, then a short course for the people running it. So the owner can make small changes without paying for each one.
And these tend to waste money:
- Project, then audit. Discovering afterwards that a tool you already paid for could have done the job.
- Course after course. Learning without applying it to a real job fades within weeks.
- An audit with nobody to act on it. A thorough report on a shelf changes nothing.
A realistic example of the first: a butcher pays for a website chat assistant to answer customer questions, and after launch finds that most questions are about collection times for online orders. The online shop's own order-confirmation email could have included the collection window all along. A one-hour look at what customers were actually asking would have cost less than the assistant's first month.
A delicatessen's route through three of the four
An illustrative delicatessen with 11 staff and a growing catering arm works through the five questions. No unapproved tools, so no audit. Staff are uneven with AI, so a course is on the list. The owner can't choose between two jobs, so a call comes first. Here's how the next three months went:
| When | Format | What happened | Owner and staff time |
|---|---|---|---|
| Week 1 | 1:1 call | Ranked the two jobs: catering enquiries (5 hours a week of the owner's time) ahead of supplier invoice matching (2 hours a week of the bookkeeper's). Decided staff needed the basics before any build | 2 hours preparing, plus the call |
| Weeks 2 to 5 | Course | Six staff took a free vendor course and practised on real tasks for a month using the deli's business AI account | About 3 hours each, then an hour a week |
| Weeks 6 to 9 | Project | A fixed-price build: catering enquiry form, AI-drafted quotes for the owner to approve, a follow-up after three days. Tested on 20 past enquiries | 6 hours of the owner's time |
| Week 13 | Check | Owner's time on enquiries down from 5 hours to about 2 a week; quotes out the same day instead of within three days | 1 hour |
The audit never happened because it didn't need to: the owner knew every subscription and every login. The course came before the project so that the staff running the catering inbox understood what the AI step was doing and trusted it enough to use it. And the invoice-matching job was deliberately left for later, once the first build had proved itself.
Your route will differ, but the method holds. Name what's missing, buy the format that fixes that, and let the next gap show itself before you pay for it.
Further reads
- Free AI Consultations: What You Get and How to Use Them Well — What a free first call can and can't settle.
- How to Run a Paid Discovery Phase Before a Full AI Project — A paid step between a call and a full project.
- AI Strategy vs AI Implementation Consultant: Which Do You Need? — Strategy or hands-on help: which kind of consultant fits.
- Is Paying for AI Help Worth It? Your Time vs an Expert's Fee — Put numbers on whichever option you're leaning towards.
- How to Audit Your AI Subscriptions and Cut Wasted Spend — A do-it-yourself subscription audit before you pay for one.
- AI Readiness Checklist: Score Your Business in 20 Minutes — Score your readiness in 20 minutes before choosing.
- Do I Need an AI Consultant? 8 Signs It's Time to Get Help — Eight signs a small business needs outside AI help, each with a test you can run this week, plus the signs that it doesn't need a consultant yet.
- Is a 1:1 AI Consultation Worth It for a Small Business? — The real cost of a 1:1 AI consultation, how many hours it must save to break even, and three cafés that get three different answers.
- How to Prepare for an AI Consultation and Leave With a Plan — The processes, volumes, timings and tool plans to gather before an AI consultation, a filled-in pre-read, the questions to ask and a written plan to leave with.
- What Is the Cheapest Way to Get Expert AI Advice? — Free help pages, forums, adoption kits and discovery calls answer most AI questions. When paid advice is worth it, and how to compare quotes.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: OpenAI Academy, Anthropic Academy course catalogue and Microsoft Learn Copilot learning paths (checked September 2026). Business examples are illustrative.