How to compare AI consultant quotes when the totals describe different jobs

Put the quotes in one table before you look at a single total. The rows are: the job in one sentence, what counts as finished, the systems it touches, whose accounts the tools live on, the monthly running cost, how changes are priced, what you hold at handover, and when you pay. A blank cell is information, because it is the part that supplier did not price. Two AI consultant quotes only compare once both describe the same job, and most of the work of comparing them is getting them to.

The totals on the cover page are the least useful number on either document. A consultant quoting less has often scoped less, and a consultant quoting more has sometimes priced a risk the other one left out. You cannot tell which from the total.

Why two AI quotes rarely describe the same job

An automation quote is a price for a guess about your business. Each consultant guessed from a different conversation, heard a different version of the problem, and filled the gaps with their own assumptions. One heard “the order entry” and priced a connection between two systems. The other heard the same words and priced the two systems, the exceptions, the reconciliation report and a monitored month afterwards.

Neither is wrong. They are answers to different questions, and the fix is to make the question the same.

Where each consultant is based matters less than it looks. This work is done remotely, so a supplier in Toronto, another in Barrie and a third in Thunder Bay are pricing the same labour and the same tools. When their totals differ, look for the reason in the scope first, and the table below is how you find it.

The comparison table

Copy this into a spreadsheet, one column per supplier. Fill it only from what the quote says in writing, not from what was said on the call.

Line	What you want to see	What a blank usually means
The job, in one sentence	A named process, start to finish, in your words	The scope will be decided after you sign
What counts as finished	A test you could run yourself on real records	Finished is whenever the supplier says it is
Systems touched	Every system by name, and which ones it writes to	The quote assumed fewer systems than you run
Whose accounts	Platform and model subscriptions on your accounts, billed to you	Changing supplier later means a rebuild
Monthly running cost	A figure or range in Canadian dollars, with the currency of each subscription	A second bill arrives after launch
Changes and exceptions	An hourly or per-change rate, and what counts as a change	Every surprise becomes a negotiation
Handover	A runbook, the credentials, and a named person who explains it	You rent the automation rather than own it
Payment	Stages tied to working pieces you can see, not to calendar dates	You pay for time rather than for results

Two blanks matter more than the others: what counts as finished, and whose accounts. The first decides whether you can ever say the job is done. The second decides whether you can ever leave.

Reading each line

The job, and what counts as finished

Rewrite each consultant’s scope in one sentence of your own. If you cannot, the scope is not specific enough to price, and the number on it is an estimate of their effort rather than a price for your outcome.

For “finished”, look for a test you could run without them: the automation processes a week of your real records, the exceptions land in a named inbox, and somebody on your staff can explain what happened to each one. A quote that says “deployed” or “live” has described an event, not a result.

Systems touched

Count the systems in each quote and compare the count to your own list. This is where the biggest price gaps usually come from, and it is often the explanation when one AI quote is several times the other. A build that reads from one system and writes to another is a different job from one that writes into your accounting package, your CRM and a spreadsheet somebody keeps by hand. Published ranges for workflow automation in Ontario run $2,500 to $12,000 on this site, and the number of systems is the main thing that moves a build along that range.

If one quote names a system the other ignores, ask the second supplier about it directly. Either they missed it or they plan to leave it manual, and you need to know which.

Whose accounts

Every automation runs on something: a workflow platform, a model provider, sometimes a telephone or messaging service. Those subscriptions should sit on accounts in your business’s name, billed to your card. A quote that bundles them into a monthly fee is not necessarily worse, but it is a different product, and it is harder to leave. The platforms also bill in different currencies and on different units, which is set out in n8n vs Make vs Zapier, so a subscription line quoted in US dollars or euros is not yet a Canadian dollar cost.

Where the process touches client records, the accounts line also decides which tools are allowed at all, for the reasons in PIPEDA compliance for AI tools in Canada. A quote that names a tool without saying where it stores personal information has left out the question that matters most.

Monthly running cost

Put every recurring number on one row, converted to Canadian dollars: platform, model usage, messaging, and any monitoring fee. Having somebody else watch it typically costs another $150 to $400 a month, as set out in the AI consulting cost breakdown for Ontario. A lower build price with a higher monthly line can cost more over the life of the build than the reverse. Do that arithmetic on the page, for both quotes, before you look at anything else.

Changes, exceptions and handover

Every automation meets a case nobody described. The quote should say how that case is priced: an hourly rate, a fixed price per change, or a stated allowance. A quote with no answer here is not cheaper. It has moved the uncertainty onto you.

For handover, ask for the runbook contents in the quote itself: every trigger, every credential and where it is kept, and what to do when it fails. If one supplier includes that and another does not, the gap is real work, and it belongs in the comparison.

Tax and payment

Compare every quote either with tax or without it, never one of each. A registered supplier charging an Ontario business for services supplied in Ontario generally charges 13% HST, so a quote stated “plus HST” and a quote stated “all in” can look further apart than they are.

For payment, stages tied to something you can see working are easier to hold a supplier to than stages tied to dates. A deposit, a payment when the first piece runs on your real records, and a final payment at handover is a shape that protects both sides.

Make the quotes comparable before you compare them

The fastest way to get comparable quotes is to send every consultant the same written brief and the same questions, then compare the answers rather than the proposals. The brief is one paragraph: the job in your words, the systems it touches, who does it today and roughly how often. The questions are the rows of the table above.

Ask for a basic AI consultant NDA and the answers in writing. If a supplier’s answer changes the total, ask for the revised total. A consultant who resists answering the same questions as everyone else is telling you how change requests will go.

If one quote is for AI agent development in Ontario and the other is for a workflow, you are not comparing two prices. You are comparing two diagnoses. On this site an agent runs $4,000 to $15,000 and a workflow $2,500 to $12,000, and the two can be quoted for the same complaint. Settle which one the job needs first, then compare prices inside that answer.

When the quotes still disagree

Sometimes the table is full, the scopes match, and the totals are still far apart. That is the point where a second opinion is worth paying for, because the disagreement is now about the diagnosis, not the paperwork.

The AI business assessment for Ontario businesses on this site is a fixed $999 over ten business days, and a build with me later has the whole fee deducted from it. It is written so another supplier can quote from it, which is the property that makes it useful in this situation: it gives every consultant on your list the same starting document. The full list of what that document should contain is in the deliverables of an AI readiness assessment. If you already have two full quotes and a clear favourite, you do not need it. Sign the one whose table has no blanks.