She has three quotes on the desk and a budget of about fifteen thousand pounds for the first six months. One is from a solo consultant, six weeks of strategy plus a small implementation. One is from an agency of about thirty people, twelve weeks of delivery against a defined scope. One is from a SaaS product, six hundred pounds per seat per month with a configuration package on top. All three are pitching the same outcome in different words. None of the three quotes makes it obvious which one fits the job she actually has.
This is the choice almost every owner runs into when they decide to put money behind AI. It is not the choice the marketing material trains you for. The marketing trains you to pick between vendors inside a category. The harder choice is between the categories themselves.
What is the actual difference between a consultant, an agency, and a SaaS tool?
A consultant is a person or small firm selling judgement plus delivery, diagnosing the right problem and often building the first version with you. An agency is a team of typically ten to fifty people selling delivery at scale against a defined brief, with their own tools and bench capacity. A SaaS tool is software you license, configure, and operate yourself, with the vendor running the product.
The three are not just different prices for the same thing. They are different products. A consultant produces a recommendation and trained internal people who can carry it. An agency produces a working system delivered to spec. A SaaS tool produces a subscription and a login. The frequent buyer mistake is treating them as substitutes when they are complements, each suited to a different kind of job.
Why does the choice matter for your business?
Because the cost of getting it wrong is rarely the fee, it is the eighteen months lost before you realise it did not work. Source Global Research’s 2025 work found AI advisory jumped from eleventh to first in client priority rankings in a year, with eighty-eight per cent of buyers now paying for it. The market is loud, and many owners default to whichever type they last had a good run with.
The win-loss patterns are clear enough in the research. Halo Tech Lab’s 2025 review of UK SME AI implementations found that the cases showing ROI in weeks to months were the ones where engagement type matched industry context, not the ones with the biggest budget. CloudZero’s 2025 cost data puts typical first AI projects at forty thousand to four hundred thousand pounds for SMEs. At those numbers, choosing the wrong category twice is the price of choosing right once.
Where will you actually meet this choice?
You meet it the day AI becomes worth real money rather than weekend experimentation. The trigger is usually a board paper saying “we need an AI strategy”, a competitor doing something visible, or a specific job that has been waiting eighteen months for a fix. YouGov’s 2026 SME survey found thirty-one per cent of UK SMEs already using AI in some form, and this is the gate they pass through.
The shape of the day is also familiar. Three to five proposals arrive over a fortnight, all using overlapping language. The agency proposal looks comprehensive and includes Gantt charts. The SaaS proposal looks cheap on a monthly basis and expensive on a twenty-four month total. The consultant proposal looks short, expensive per week, and slightly vague on deliverables. The mistake is comparing them on price per page rather than on what each one actually produces at the end.
Which engagement type fits which job?
Four questions decide it. How unusual is the use case, has it been solved before in firms like yours? How much hand-holding does the buyer need, can someone internally run a brief and govern scope? What do you want to own at the end, a working system, or a built capability? And what is the total cost across twelve to twenty-four months, not just the first invoice?
Use the answers as filters. Unusual use case plus owner-needs-to-learn-alongside plus judgement-heavy decisions ahead, that is the consultant case. McKinsey’s work identifies five strategy roles a consultant plays, researcher, interpreter, thought partner, simulator, communicator, none of which a SaaS tool replicates. Well-trodden problem plus internal capacity is the limit plus clear requirements, that is the agency case. BCG’s Deploy framework fits here, you know what good looks like and the constraint is execution, not direction.
Commodity use case plus willingness to adapt to the tool’s workflow plus low switching cost if it does not work, that is the SaaS case. High Alpha’s 2024 SaaS benchmarks show vertical SaaS achieving net revenue retention of 120 per cent or higher against 100 to 110 per cent for horizontal tools, which is to say industry-specific SaaS is genuinely good at commodity jobs in its industry. The trap is using SaaS for a non-commodity job, where Blissfully research finds roughly thirty-seven per cent of SaaS spend is wasted through poor adoption. Budget gates the choice too, under thirty thousand pounds over twelve months and SaaS is the only realistic path, regardless of fit.
What to ask before you decide
Four questions clear the fog before you sign. Ask the consultant which off-the-shelf tools they considered and rejected for your case, and why. Ask the agency to write the statement of work in plain English, three things in, three things out, and what happens when scope creeps in week four. Ask the SaaS vendor for the total cost over twenty-four months, including pricing after the first-year discount expires.
Then ask yourself the harder question. Six months from the end of this engagement, what does my business actually own, and what does it depend on? If the answer is “a working system and a team that knows how to run it”, the engagement is shaped right. If the answer is “a subscription and a hope that the vendor stays in business”, you might still buy, but you buy with your eyes open. The four discriminators are not a scoring rubric, they are a way to stop the default trap. Whichever type of provider you last had a good experience with is the type you will reach for again. The job in front of you may need something else.
A useful sanity test is to write the answer to the four discriminators on a single page before any of the three vendors hears back. If you cannot fill the page from the information already on your desk, that is a signal you are framing too fast and need another round of clarifying questions on your side, not on theirs. Book a conversation if you would like a second pair of eyes on the three quotes before you sign.
