Skip to content

ArticleChoosing an AI Partner

AI implementation cost: what you pay to install it and to run it.

Most AI budgets count the build and forget the years after it. Here are the five cost lines of an AI implementation, why quotes for the same work vary so much, and a one-workflow budget sheet you can fill in with your own numbers.

The AI implementation cost on a quote is usually one line of a longer bill. The build is the part everyone prices. Discovery, integration, the hours your team spends adopting the new way of working, and the monthly cost of keeping it running arrive later, often from different budgets. If you are comparing providers, read this next to what AI consulting costs, which covers how each kind of provider prices the work.

This article lays out every line, so you can budget for the whole thing before you sign for part of it.

The five cost lines of an AI implementation.

  1. Discovery.

    Working out which workflow to fix first and what it should return. Some providers sell it as a paid project; others fold it into the build; a free diagnostic covers it with no invoice at all.
  2. Build or install.

    Building a system from scratch, or installing one that already exists and configuring it to your systems and approval rules. Building from scratch costs more and takes longer, because the provider is inventing the solution as well as fitting it.
  3. Connecting your systems.

    Connecting the system to what it has to read and write: the inbox, the ERP, the CRM, the accounting system. Access, permissions and testing sit here, and so do the surprises.
  4. Adoption and training.

    The hours your team spends learning the new flow, reviewing exceptions and changing habits in the first weeks. It rarely appears on a quote, and it is real money.
  5. Running costs.

    Monitoring, fixes, updates when a model or a connected system changes, model usage fees, and a person accountable for the result every month for as long as it runs.

Why published price ranges vary so much.

Two quotes for "automate our invoice processing" can differ by a wide margin and both be honest. The drivers are the same every time: whether the system is built from scratch or installed from something that already exists, how many systems it has to connect to, how much work the data needs, and who runs it afterward. Each quote prices a different share of the five lines above, under a different model.

How it is pricedWhat the quote usually coversWhat it tends to leave out
Hourly or daily ratesThe time spent on the engagementThe result, and everything after the engagement ends
A fixed project feeA defined build, to a written scopeChanges, support and model updates after handoff
Software licensesThe tool, per seat, per monthSetup, adoption and fitting it to your workflow
An in-house hireOne person's salary and timeRecruiting, management and the gap if they leave
A fixed monthly subscriptionThe install plus running it, reported in dollarsCustom work outside the provider's library, priced separately

Before comparing numbers, put each quote against the five lines and mark what it includes. The cheapest quote is often the one that stops earliest. The five ways to get AI working are compared side by side in the partner guide.

Custom build or a pre-built AI system.

A custom build fits when the workflow or its data is unusual, or when the system itself is part of what you sell. You pay for the design as well as the build, and you own the upkeep of everything that was invented for you.

A pre-built AI system fits when the workflow is common: quotes, order entry, invoices, lead response, reporting. The design work was paid for once, so you pay to configure it to your systems and rules. The trade-off is fit. If the system does not match how your business works, it is the wrong choice, and a good provider will say so before you sign.

The running cost most quotes leave out.

An AI system is closer to a machine on the shop floor than a report on a shelf. It needs someone watching it, and it needs maintenance. Models are retired and replaced. Vendors change their prices. The ERP gets an update and a field moves. Every one of those events lands on someone's desk.

Budget for the second year before you sign for the first.

Ask every provider the same three questions. Who runs this after launch? What does that cost each month? What happens when a model changes or gets more expensive? If the answer to the first question is "your team", add that team's time to the running line. The guide on running AI after launch covers this line in detail.

Budget one workflow for one year.

Fill this in for the one workflow you would fix first. Ask each provider to fill in their lines, and fill in the adoption line yourself.

LineWhat to countYour amount
DiscoveryOne time. The diagnostic, assessment or scoping work.$
Build or installOne time. Building or configuring the system.$
Connecting your systemsOne time. Connections, permissions, testing.$
Adoption and trainingTeam hours in the first weeks, times the loaded cost of an hour.$
Running costs, per monthMonitoring, updates, usage fees, the accountable person.$
Year oneThe one-time lines plus twelve months of running it.$
Year twoTwelve months of running it.$

Set the year-one total against what the workflow costs you today, and year two against what it saves.

The adoption line is the one owners underestimate most, so here is how to size it.

Sizing the adoption lineIllustrative math
People who touch the workflow5
Hours each in the first month6
Loaded cost of an hour$45
Adoption, first month$1,350

Your own team size and hourly cost go here. Most of the hours are reviewing exceptions while the system settles in.

What StrikeArc charges.

StrikeArc publishes its prices, so you can fill in the sheet before the first call. StrikeArc Scout, the Intake & Diagnostic, covers discovery and is free; the readout is yours to keep either way. The StrikeArc PowerPlays Library subscription is $7,500 a month, including one play installed per quarter, plus run, optimization and ROI reporting on everything live. A one-time install of an existing play is $20,000 per play, installed and handed over. When nothing in our library fits, Custom Discovery is $10,000, then a Custom Build and a Custom Subscription priced from the Discovery findings. See the prices in full, or how StrikeArc compares with the other ways to get AI working.

Before you sign.

Whoever you choose, get four answers in writing before you commit.

  1. What is in the price?

    Mark each of the five cost lines as included, extra or yours to carry.
  2. What runs monthly, and what does it cost?

    Model usage, monitoring and updates, stated as a monthly figure.
  3. Who runs it after launch?

    A named role on their side, or a person on yours whose time you now have to budget.
  4. What do you own if you stop?

    The configuration, the data and the documentation, and what happens to each on the day the contract ends.

More on weighing providers against each other is in the guide to choosing an AI implementation partner.

Keep reading.

How to get started

Get the full cost on one page before you commit.

Book an intro call

Thirty minutes with the exec who owns the problem. We will tell you straight whether a play fits and what it costs.