An agent install costs a fixed setup fee plus a monthly fee. Most of ours cost $1,500 to $5,000 to set up and $1,500 to $3,000 a month (October 2026). A custom AI build is priced on engineering time. The average AI project on Clutch costs $120,595 over 10 months. Choose a custom build when your data must stay in-house.

What does a custom AI build cost in 2026?

Clutch's AI pricing page puts the average AI project at $120,595 over a typical 10 months. Clutch updated the page on 21 September 2026 from verified client reviews. The most common project size is $10,000 to $49,999.

A buyer guide from Winder.ai, published 24 June 2026, gives one firm's estimate for each stage:

  • Proof of concept: £15,000 to £40,000, 2 to 4 weeks.
  • Pilot or MVP: £40,000 to £120,000, 6 to 10 weeks.
  • Mid-market production build: £120,000 to £400,000, 4 to 6 months.
  • Enterprise or regulated production build: £400,000 to £1.5M or more, 6 to 12 months.
  • Running costs after launch: 15% to 25% of the build cost each year.

UK public-sector rate cards on G-Cloud 15 show the day rates behind those totals (January 2026):

  • One MLOps supplier charges £950 a day for an engineer and £1,750 for a strategic lead. Its 8-week example of 136 days costs £154,000.
  • A large consultancy lists a senior machine learning engineer at £1,530 a day onshore and £700 offshore.

What does an agent install cost?

Each agent team has a fixed price, set before you start. Our prices, confirmed 16 October 2026:

  • Sales Intelligence Team: $1,500 setup, $1,500 a month.
  • Content Team: $1,500 setup, $2,000 a month.
  • Community Team: $2,500 setup, $1,500 a month.
  • Institutional BD Team: $5,000 setup, $3,000 a month.
  • SDR Team: runs your campaign inside First Pipeline, $15,000 for 90 days, with no setup fee and no fee per meeting.

The first year of a Sales Intelligence Team costs $19,500: $1,500 setup plus 12 months at $1,500. An Institutional BD Team costs $41,000 over the same year. These are about a sixth and about a third of the Clutch average project.

The price is lower because the engineering already exists. An install sets up a built system with your inputs: your ideal customer profile, documents, voice samples and Telegram group. A custom build pays engineers to scope, build, test and govern a new system.

How do the two compare?

Agent install

  • Ongoing cost: a fixed monthly fee.
  • First output: a weekly digest from Sales Intelligence. Outbound sends after a 14-day mailbox warmup.
  • Usual buyer: a founder, CEO or sales leader at a company of 8 to 500 staff.

Custom AI build

  • Ongoing cost: 15% to 25% of the build cost a year (Winder.ai estimate).
  • First output: 2 to 4 weeks for a proof of concept, 4 to 12 months for production (Winder.ai estimate).
  • Usual buyer: a CIO, Head of Data or Chief AI Officer, often with procurement.

When is a custom build the right choice?

Choose a custom build when the job sits outside sales and marketing, or the data cannot leave your control. Three cases point to it:

  1. Regulated data must stay in-house. Examples are patient records, card data or bank customer data under a policy that forbids third-party model APIs. Our agents run on hosted model APIs, so they do not fit this case.
  2. You need deep custom machine learning, such as fine-tuned models or models trained on proprietary data. This work needs ML engineers for months.
  3. The job is an internal operation, such as claims handling, document review or finance workflows. A fixed agent team for pipeline has no version of these jobs.

If one of these cases fits, budget for the proof of concept first. Ask for the running cost in writing before the build starts.

Why does the market split this way?

Gartner forecasts $576.5 billion of AI services spending in 2026, and $29.2 billion on AI agents and assistants (techstrong.ai, 21 September 2026). Agent spending grows 77.3% this year in the forecast (CIO.com, 23 September 2026).

A company of 50 staff that wants more qualified meetings rarely needs a 10-month project. It needs a working agent on a known job, at a price it can approve without procurement. Larger firms with regulated data and internal processes need the custom build.

For the cost lines behind a sales agent, read what a sales AI agent costs in 2026. To judge whether an agent can take over SDR work, read can AI agents replace SDRs.

Which one should you buy?

Answer these five questions before you ask for a quote:

  • Does the job sit in sales, marketing, community or BD, with a fixed output?
  • Can the data pass through a hosted model API under your policies?
  • Do you need results this quarter, before a 4 to 12 month build could finish?
  • Is your first-year budget for this job under $50,000?
  • Will someone on your team own the inputs, such as the documents and the replies?

Four or five ticks point to an agent install. Two or fewer, with regulated data or an internal operations job, point to a custom build.

To see the buyers our agents would find for your pipeline, start with five named buyers, free. Each one comes with the reason they fit now.