← Back to blog
2026-09-03·AI Strategy·7 min read

How to Read an Enterprise AI or Data Quote

A big AI or data quote usually arrives as one number and a list of line items you rarely have to buy. The number gets the attention, but each line item is a decision the vendor made for you, and each one can be questioned. Here is what the common lines mean, when each is fair, and what to ask.

In short: Read a quote line by line instead of starting from the total. The usual lines are discovery and consulting, platform licences, the build itself, data migration, training, and annual support. Every line can be legitimate; whether it is depends on whether it is scoped to your size and your problem. Fair Singapore rates for the build work are covered in our cost guide; this post is about everything wrapped around the build.

What are the line items, and when is each one fair?

Line itemWhat it isFair whenAsk this
Discovery / consultingPaid scoping before the buildIt is short, fixed-price, and ends in a document you own"Can we do discovery alone first, and keep the scope?"
Platform licencesSoftware the solution runs onThe platform is needed at your scale"What does this cost in year two, and per person?"
The buildThe actual development workScope is fixed and matched to your problem tier"Which of the three kinds of AI build is this?"
Data migrationMoving your existing data inYour data genuinely moves systems"What happens if we keep the current systems and integrate instead?"
TrainingTeaching your team the resultIt happens after go-live, on real tasks"Is there support when questions come up in month two?"
Support / maintenanceKeeping it running, yearlyIt is quoted up front as a yearly figure"What is included, and what costs extra per incident?"

Two of these deserve special attention in SME quotes, because they are the lines most often priced for a larger company.

Why do discovery and consulting lines grow so large?

Discovery is real work, and a short paid scoping phase usually saves more than it costs; it is exactly what an audit is. The line becomes a problem when it is open-ended, priced by the day with no end date, or bundled so you cannot take the findings elsewhere.

A fair discovery line is short, fixed, and produces something you own: a scope document with prices that any competent builder could execute. If the findings are only usable when you buy the build from the same vendor, the document is not independent scoping you can take elsewhere, and the line should be challenged.

What should you check on platform and licence lines?

Three things:

  • Scale fit. Licences are often tiered for user counts and data volumes an SME will not reach. Ask what the tier assumes, and what the smallest viable tier costs.
  • Year-two cost. The first year is sometimes discounted into the build price. The renewal is the real number; get it in writing.
  • Whether a platform is needed at all. Many SME problems are integration work on tools you already own, where the licence line should be small or absent. If the platform is the biggest line, that is the first thing to challenge, and the build-versus-buy question is the frame for challenging it.

What is missing from most quotes?

The two lines that determine whether the project pays off usually are not priced at all:

  • Adoption. Who moves your team's workflows into the new system and stays until usage holds? Unpriced adoption is how good software ends up unused. If the quote has no answer, the real cost of the project is higher than the total at the bottom.
  • Exit. Who owns the code and the data, and what does leaving cost? Per-seat pricing, vendor-locked platforms, and proprietary data formats all raise the cost of leaving later. Ask the ownership question before signing, while you can still walk away.

How do you check the whole quote?

Four steps, doable in one afternoon:

  1. Match the total to your problem tier. A manual-workflow problem priced like a platform build has a scale mismatch somewhere in the lines; the cost guide has the Singapore ranges per tier.
  2. Ask for the quote split. Discovery, build, licences, and yearly running cost as separate numbers you could in principle buy separately. If a vendor will not split the quote, treat that as a warning sign.
  3. Get year-two in writing. Licences, support, and any per-use fees, as a single yearly figure.
  4. Price the alternative. A second opinion on scope costs far less than the delta between two quote tiers. A two-week audit (from SGD 4,000) produces an independent scope with prices, which either confirms the quote or gives you the numbers to negotiate it.

What to do next

Before responding to the quote on your desk:

  • Mark each line item against the table above: fair, question, or challenge.
  • Write down the two unpriced items, adoption and exit, and ask the vendor for both.
  • Get the year-two number in writing.

A vendor with a fair quote will answer all three quickly.

Not sure it's worth it?

A jinq AI Audit (two weeks, remote, from SGD 4,000) works as an independent second opinion on a quote: we scope the same problem at your size and hand you the numbers, whether or not we do the build. If the quote you have is fair, we will say so. If you want the work done and adopted, a Fractional AI Officer (from SGD 7,500 a month) can run it one to two days a week.