Every Newcastle business owner we speak to has the same first question. What does this cost.
The answer most agencies give is "it depends." That is technically true and practically useless. Here is how we actually price production AI software in 2026, why we run it as a retainer, and what drives the number on your specific engagement.
We operate on a retainer, not a per-project basis
Production AI infrastructure is not a one-off build. The first workflow proves the value. The second, third, and fourth compound on top of it. Models move forward, integrations drift as your upstream systems change, compliance regimes tighten, and the business case for new automations keeps surfacing as the team sees what the first ones did.
A project-by-project engagement model does not fit that shape. A retainer does. We commit an engineering function to the business, and the business gets continuous build, maintenance, and new capability shipped without renegotiating scope every quarter.
The two retainer tiers
There are two entry points, both monthly, both on a 12-month minimum term.
Existing systems retainer: from £4,000 per month
For businesses that already have AI or automation running and need it properly owned. Stewardship of what is live, remediation of no-code or prototype workflows that have become load-bearing, and a steady cadence of new workflows on top of the existing base.
This is typically the right starting point if you have something in production already and it is costing more in workarounds than it saves.
New systems retainer: from £6,000 per month
For businesses starting from zero. Full greenfield: discovery, architecture, first workflow in production inside six weeks, ongoing build from there. Everything is engineered for your infrastructure from the outset rather than retrofitted onto someone else's platform.
This is the right starting point if you are committing to production AI for the first time and want it built properly.
Scope-dependent, scaling up from the floor
The numbers above are the floors, not the ceiling. Retainers scale with scope. A single-site operator automating three or four workflows sits near the floor. A multi-site group with regulated data, deep integrations across a legacy ERP, and parallel workstreams across finance, ops, and compliance sits well above it. Exact pricing is set on consultation, against a written scope, so you know what you are paying for and what is in each month.
What drives cost within a retainer
The retainer model means your budget buys engineering hours, not licence seats. What those hours are spent on depends on four variables.
Integration complexity. Reading from and writing to your ERP, CRM, finance system, or bespoke platform is the most time-variable part of any build. Modern REST APIs are fast. Legacy systems requiring RPA, screen-scraping, or bespoke adapters take longer. A finance automation into Xero is cheap to integrate. The same automation into a bespoke 1990s accounts package is not.
Data quality. Clean structured data is cheap to work with. Noisy unstructured data (handwritten forms, inconsistent PDFs, voice transcripts, free-text shift reports) needs more pipeline engineering to become reliable. The AI part is rarely the bottleneck; the pre-processing and validation around it usually is.
Compliance requirements. Regulated sectors (finance, legal, healthcare, defence, education holding sensitive personal data) need audit trails, private AI on-premise, tighter data handling, and evidence on demand for a supervisor visit. That adds engineering time, and it is not optional. It is also why we are ISO 27001:2022 certified and Cyber Essentials accredited: regulated clients can only buy from vendors who have done this groundwork properly.
Scope and cadence. A retainer with one concurrent workstream moves more slowly and costs less than a retainer with three running in parallel. Clients dial this up and down as their roadmap demands.
Why Newcastle
This is not a cost arbitrage. Newcastle has an unusually deep engineering talent pool: Northumbria, Newcastle University, Durham, plus Sage and the Nissan tier-one supply chain have trained generations of enterprise software engineers in the region. The density of serious engineering here is the reason we build from Newcastle, not a discount positioning.
The output is software that can hold up in production for years, with the governance and documentation to match. Clients pay a premium rate for premium engineering. They get a system they own outright, built by a team that will still be here in five years.
What it does not include
No per-seat software licences on work we build. No cloud LLM subscription that you pay us every month. We build software you own, running on infrastructure you control, with model choices you can swap without renegotiating.
Hosting and model API costs sit outside the retainer and are paid directly by you to your cloud and model providers. These typically run £100 to £500 per month at SMB volumes, billed transparently to your own accounts.
The structural reason privately-deployed AI is cheaper over a multi-year horizon than SaaS equivalents is laid out in on-device AI architecture cost vs cloud LLMs.
What the return looks like
Retainer engagements typically pay back inside the 12-month term in administrative time recovered and new revenue unlocked. The range of outcomes we ship across clients is wide: a poker operator recovered 18% of revenue through real-time fraud detection, an accounting firm reduced a 12-person admin team to 2 senior reviewers with 10 redeployed to sales, a consultancy compressed a 700-hour process to 30 minutes, and a higher education client went from 22 days to 4 hours on DPO and SAR work.
Your specific return depends on which processes you automate first and how much manual admin sits behind them today. See the cost of not automating for the calculation method.
Getting started
Exact pricing is set on consultation. After a 30-minute discovery call we send a written scope and a fixed monthly number within a week. No multi-round procurement cycles, no hidden change fees, no per-seat surprises at renewal.