Start with referrals from businesses like yours, then the partner directories run by the tools you already use (Zapier, Make, HubSpot and Microsoft all list certified partners), then freelance marketplaces for small fixed jobs. Shortlist three, send each the same short brief about one real problem, and judge them on how they answer it.
Where you look matters less than what you do with the names. Every channel turns up good and bad consultants; the difference is the filter you apply. A referral that comes with a specific result, a directory listing with verified reviews, or a freelancer with finished jobs like yours is a starting point. None of them is a reason to skip the brief.
Match the channel to the size of the job
Different channels suit different kinds of work, and the most common mistake is using a channel built for one kind of job to find help with another. Decide which of these you need before you search:
| What you need | Best first channel | Watch for |
|---|---|---|
| One defined automation in a tool you already use, such as a Zapier or Make workflow | The tool's partner directory, or a freelance marketplace | A well-built job that nobody in your team understands afterwards |
| AI features switched on and set up in a CRM, helpdesk or Microsoft 365 | That vendor's partner directory | Partners who earn from the vendor recommending upgrades you don't need |
| Working out where AI would pay off across the business, before any tool is chosen | Referrals to independent consultants | Anyone tied to one vendor, who will find that vendor's answer |
| Ongoing help after a first project | Whoever did the first project well, or a referral | Retainers with no defined deliverables |
The third row matters most. If you don't yet know which tool you need, a vendor's partner directory is the wrong place to start, because every partner in it has a commercial reason to recommend that vendor. That's not dishonest; it's how the programmes work, and checking whether a consultant earns commission on the tools they recommend explains what to ask. For the broader question of which type of provider suits you, AI consultant vs agency vs freelancer compares them.
Referrals from businesses one step ahead of you
Time: about an hour of messages, then a week of waiting for replies.
The best referrals come from organisations a little bigger or a little further along than yours, in a similar line of work: they've faced your problem recently, and their consultant worked at your scale. Ask peers in your sector, your accountant and your IT support company, other members of any business network you belong to, and, for charities and social enterprises, your funders, who often know which advisers their grantees rate.
A vague request gets vague names. Ask for specifics, so the referral arrives with evidence attached:
Hi [first name], we're looking for help using AI to cut the admin in our volunteer scheduling and funder reports. Have you worked with anyone on something similar? If so, I'd love to know what they actually set up, whether it's still running, and what you'd do differently if you hired them again.
"Whether it's still running" is the most useful question in that message. A consultant whose work is still in daily use a year later has passed the test that matters most, and one whose automations were quietly switched off after a month has told you something too.
Be wary of one thing: a warm referral can make you skip the checks you'd apply to a stranger. Treat a referred name as a candidate to test, not a decision already made.
Partner directories run by the tools you already use
Time: one to two hours to search and read profiles.
If the job sits inside a tool you've already chosen, its partner directory is often the quickest route to someone who knows it well. The main ones, as they stand now:
- Zapier's Solution Partner Directory lets you filter by the tools a partner most commonly works with, the industries and company sizes they mainly serve, and whether they're available, and sort by tier and review count. Partners progress through tiers that start at the Certified Zapier Expert level. Zapier also offers to match you with partners.
- Make's partners directory filters by language and specialisation and ranks partners in Platinum, Gold and Silver tiers, with a "Match me" option. For smaller jobs, Make points people to the "Hire a pro" section of its community forum.
- HubSpot's Solutions Directory lists Solutions Partners with client reviews and tiers running from Gold up to Elite. It suits CRM setup, including the AI features in HubSpot's paid hubs.
- The Microsoft 365 Copilot Partner Directory lists partners for readiness assessments, deployment, adoption and change management, agent security and governance, and AI advisory work.
Tiers and certifications tell you a partner has passed the vendor's exams and has a track record with the vendor's product. They don't tell you the partner understands small organisations, so filter for company size where the directory allows it and read the reviews for clients like you. Remember the commercial link too: HubSpot, for example, pays Solutions Partners a revenue share of 20% of net revenue on qualifying sales for up to 36 months. For platform-specific hiring, see the tutorials in the reading list below.
Freelance marketplaces: fees, filters and fixed-price jobs
Time: an hour to post a job, then two to three hours to read proposals.
Marketplaces such as Upwork and Fiverr work best for a job you can describe precisely: "build a Make scenario that turns our booking form into a calendar entry and a confirmation email" rather than "help us with AI". They're weakest for the thinking-through stage, because the format rewards quick, cheap proposals over questions.
Know the platform's own charges before comparing prices. Upwork adds a marketplace fee on top of what you pay the freelancer, plus a one-off fee for each new contract, and Fiverr adds a service fee at checkout. The rates depend on the plan and change from time to time, so look up each platform's current client fees and add them to the freelancer's price before comparing it with a direct quote. On a job of a few thousand dollars, the platform's share can be enough to change which quote is cheaper.
The profiles to shortlist show finished work like yours, described specifically, with reviews from clients who say what was built. A profile full of five-star reviews for "great communication" tells you less than one detailed review of a job like yours. How to hire an AI automation freelancer on Upwork or Fiverr covers job posts, proposals and milestones in detail.
Review sites and LinkedIn: useful, with caveats
Review directories. Clutch, which lists agencies and consultancies, checks reviews in three ways: the reviewer signs in through LinkedIn, Google or a company email address; its software compares the reviewer's online footprint with what they entered; and an editor checks each submission before it's published. When a firm submits clients as references, Clutch's analysts may interview those clients by phone. That makes the reviews real, but not random: the firm usually chooses which clients to put forward. Read them for detail about how problems were handled, not for the star rating. Check the size, too. Clutch listings show each firm's minimum project size, and in its AI consulting directory these run from $5,000 or more to $50,000 or more; its AI pricing guide, updated on 21 September 2026, puts the AI development projects reviewed on the site at $10,000 to $49,999. Many of those firms are built for budgets well above a first small-business project, so look at the minimum before you get in touch.
LinkedIn. LinkedIn's Services Marketplace lets you request a proposal from providers without paying for InMail, and a search for a specific skill plus your sector ("Make automation charity", "HubSpot AI setup membership") finds people who describe the work you need. Judge them by what they show: write-ups of real projects, before-and-after descriptions, named tools. Follower counts and confident posts about the future of AI tell you about marketing skill, which is a different thing.
A community interest company builds a shortlist of three
The organisation below is invented; every step it takes is one you can copy. A community interest company (a social enterprise whose profits go back into its community purpose) runs a community café and a volunteer programme with eight staff and about 60 volunteers. Two jobs eat the most time: scheduling volunteers each week, and assembling quarterly reports for three funders. The company already uses Make for a couple of small automations and Google Workspace for everything else.
Week one: collect names. The operations lead asked a funder and two peer organisations for referrals (two names came back), searched Make's partners directory for partners working in the team's language with a non-profit or reporting specialism (three names), posted a scoped job on Upwork (two credible proposals among a dozen) and searched LinkedIn (two names). Nine candidates, about five hours of work in total.
First filter, fifteen minutes each. Did the website or profile show work for organisations under 50 people? Did it name the tools they build with? Was there any evidence of work still in use? Three dropped out: one only showed enterprise clients, one described "AI transformation" without naming a single tool or result, and one had no examples at all. Six remained.
Second filter: the brief. All six received the same one-page brief (below) and were given ten days to reply. Five replied. Two replies were generic, and three engaged with the actual problem. Those three were the shortlist: one referral, one Make partner and one Upwork freelancer.
The operations lead kept a simple log as the names came in, which made the filtering quick and the reasons easy to explain to the directors afterwards:
| Candidate | Channel | First filter | Brief | Outcome |
|---|---|---|---|---|
| 1 | Referral from a funder | Pass: two charity clients, names Make and Google Sheets | Specific questions, fixed-price first step | Shortlisted |
| 2 | Referral from a peer | Pass | No reply by the deadline | Dropped |
| 3 | Make directory, Gold tier | Pass: reporting work for small non-profits | Questions, a reasoned first step, disclosed commission | Shortlisted |
| 4 | Make directory, Platinum tier | Fail: enterprise clients only | Not sent | Dropped |
| 5 | Make directory, Silver tier | Pass | Generic reply offering a call | Dropped |
| 6 | Upwork proposal | Pass: three reviewed jobs like ours | Clear questions and a small fixed job | Shortlisted |
| 7 | Upwork proposal | Fail: no examples | Not sent | Dropped |
| 8 | LinkedIn search | Pass | Generic reply proposing a custom app | Dropped |
| 9 | LinkedIn search | Fail: "AI transformation", no tools or results named | Not sent | Dropped |
Tier and channel didn't predict the outcome. The Platinum partner was the highest tier in the directory and failed the first filter on client size; the Upwork freelancer, the cheapest-looking channel, made the shortlist on the strength of three reviewed jobs that matched. Each channel produced one good candidate. The brief did the choosing.
The decision came after a paid discovery session with the top two, a small, fixed-price piece of work that showed how each one thinks before any bigger commitment. The whole search took about three weeks and roughly eleven hours of the operations lead's time, much of it reading. Next, the shortlist went through the checks in how to vet an AI consultant's case studies and references.
The first message to send, and replies worth reading twice
The brief is the most useful filter you have, because it asks every candidate the same question and lets you compare their answers directly. Keep it to a page:
Subject: Brief: reducing admin in volunteer scheduling and reporting
Who we are: a community interest company, 8 staff, 60 volunteers,
running a community cafe and volunteer programme.
The problem: scheduling volunteers takes about 4 hours a week by
email and a spreadsheet. Quarterly reports for 3 funders take about
40 hours a quarter, mostly copying figures and stories from notes.
Tools we use: Google Workspace, Make (two small scenarios), a
volunteer spreadsheet. No developer on staff.
Constraints: volunteer and participant data is personal; some is
sensitive. Budget for a first phase is modest and must be fixed.
Please reply by [date] with:
1. The questions you'd need answered before proposing anything.
2. What you'd suggest as a small first step, and why that one.
3. How you'd price that first step.
4. Any partnerships or commissions with the tools you might suggest.Two illustrative replies show the difference. The first:
Thanks for reaching out! We help organisations realise the full potential of AI with bespoke agentic solutions. Our proven framework has delivered results for clients across every sector. Let's book a discovery call to explore how we can transform your operations.
The second:
Two questions first: how many volunteers change their availability in a typical week, and do the three funders want the same figures in different formats, or different figures? I'd start with scheduling, because the reports depend on data you don't yet capture cleanly. A first step would be a form-to-sheet flow in Make with an AI step that drafts the weekly rota for you to approve, priced as a fixed fee after a one-hour call. I'm a Make partner and earn a referral commission if you move to a paid Make plan through my link; you're already on Make, so that's unlikely to arise.
The second reply asks real questions, argues for a sequence, names the tools, explains how the price will be set and discloses a commission unprompted. The first could have been sent to anyone. Score every reply on those five points: questions asked, a reasoned first step, named tools, a pricing approach, and disclosure. I offer a 1:1 AI implementation consultation myself, so hold me to exactly the same five points. For a longer list of questions to put to your shortlist, 20 questions to ask an AI consultant first is a good next step.
Channels that look promising and usually disappoint
A few sources produce plenty of names and few good fits:
- Unsolicited messages. A cold message promising AI agents that will multiply your revenue has already shown you how its sender works: generic claims, no questions about your business. The occasional good consultant does cold outreach, but you'll find them faster through the brief.
- Free "AI audits" from firms that sell one platform. Some are genuinely useful, but an audit that ends by recommending the auditor's own subscription wasn't an independent assessment. Ask before booking what the audit usually recommends and whether the firm earns from it.
- Software vendors' own sales teams for strategy questions. They know their product thoroughly and are the right people to ask how it works. They're the wrong people to ask whether you need it.
- Ranked "top AI consultants" lists. Many are paid placements or affiliate pages. If a list doesn't say how firms were chosen, treat it as advertising.
Here's how these channels catch people. Go back to the operations lead from the example above and change one thing: in a busy first week, the lead replies to a well-written LinkedIn message and agrees to a call before sending any brief. The call is pleasant and ends with a proposal for a custom-built scheduling app. Only when the Make partner's reply to the brief suggests a form and a spreadsheet flow does the lead realise the custom build was never necessary. The brief exists to prevent exactly this, so send it before any call, to everyone, including the people who approached you first. If a candidate won't answer a one-page brief, they've answered the question you were really asking.
Further reads
- How to Hire a Certified Zapier Expert and What They Charge — What a certified Zapier expert does and how they price.
- How to Hire a HubSpot or CRM Consultant to Set Up AI Features — Hiring a HubSpot or CRM partner for AI features.
- How to Hire a Microsoft 365 Partner for a Copilot Rollout — Choosing a Microsoft partner for a Copilot rollout.
- How Much Does an AI Consultant Cost for a Small Business? — What AI consultants charge and how quotes are built.
- How to Run a Paid Discovery Phase Before a Full AI Project — A low-risk way to test a consultant before a full project.
- How to Prepare for an AI Consultation and Leave With a Plan — Get the most from the first call once you've chosen.
- How to Write a Request for Proposal for an AI Project — Eight sections every AI project RFP needs, a members' club's RFP filled in, a pricing table that makes replies comparable, and a scoring matrix.
- AI Consultant Red Flags: 12 Warning Signs to Walk Away From — Twelve warning signs when hiring an AI consultant, each with an example, the question to ask, its innocent version and a scored two-proposal comparison.
- 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: Zapier Solution Partner Directory and partner programme pages; Make partners directory and affiliate/partner pages; HubSpot Solutions Directory and Solutions Partner Program policies; Microsoft 365 Copilot Partner Directory; Clutch help pages on review verification, its AI consulting directory and its AI pricing guide (updated 21 September 2026); LinkedIn help pages on the Services Marketplace.