What drives the cost of an enterprise AI project?
The model is rarely the biggest cost in an enterprise AI project. Integration with core systems, data work, controls and the cost of running it after launch usually decide the budget. Here is how to estimate each before you sign.
Intelligent systems2 min read
The cost of an enterprise AI project is set less by the model than by four other things: the data work to make inputs usable, the integration that puts results into the systems where people decide, the controls a regulator or auditor will expect, and the cost of running it once it is live. A quote that prices the model and treats the rest as assumptions will grow.
Data preparation
Data that looks complete in a report often is not complete per record. Codes change over the years, fields are reused for new purposes, and joins between systems break on customer identifiers. The cost depends on how many source systems the model reads and how much history it needs. Sample the live data during scoping, not after the contract, and price the cleanup you find.
Integration with the systems of record
In our experience this is usually the largest item and the one most often left out. Putting a score into a core banking screen, a recommendation into an ERP planning view, or a draft into a CRM record means work in systems that have their own release cycles, vendors and test environments. The number of systems touched, and whether each one exposes an API or needs a vendor change, drives the cost more than any choice of model.
Controls
A model that affects customers needs decision logs, a human review step where risk demands it, monitoring for drift, a way to switch it off, and documentation an auditor can read. In banking, RBI's Governor has asked for human oversight and the ability to explain decisions that affect customers. Price them as part of the first release; adding them later costs more because the integration has to be reopened.
Running it after launch
Models are retrained, thresholds tuned, prompts revised, and every change tested. Language models add usage costs that rise with volume, so estimate them at production volumes, not pilot volumes. Someone also has to watch the monitoring and respond when it fires. If this cost is not in the business case, the project may succeed technically and still be switched off a year later.
How to get a reliable number
Ask for the estimate broken into these four parts plus the build, with the assumptions behind each. Ask what happens to the price if the data turns out worse than sampled. And fix the scope with a short paid assessment before a large commitment: ours is a fixed fee over four weeks and ends with a costed plan the client keeps, whether or not we build it.
Questions
- What is the biggest cost in an enterprise AI project?
- Usually integration: putting the model's output into the core banking, ERP, CRM or clinical system where decisions are made, often across several systems with their own vendors and release cycles.
- What ongoing costs does an AI system have after launch?
- Retraining and testing, monitoring for drift, human review where required, usage costs for language models at production volume, and staff to respond when monitoring raises an alert.
- How can we get a reliable estimate before committing?
- Ask for the estimate split into data work, integration, controls, running costs and the build, with assumptions for each, and fix the scope with a short paid assessment first.
More on intelligent systems
All writingNovember 2026
CERT-In's AI security audit guidelines: what an auditor will check
CERT-In's AI Security Auditing Guidelines, dated November 4, 2026, sort AI systems into four risk tiers and audit them across nine domains. For AI agents, auditors will test whether an agent can exceed its permissions or bypass approval. Here is what to prepare.
October 2026
How do you test an LLM before it goes live in a regulated business?
Test a language model the way an auditor will question it: on your own cases, against written pass criteria, with attacks on its permissions and data, and with a record of every result. RBI's Governor has asked banks to red-team and stress-test AI before deployment and periodically after; five test sets are where that starts.
October 2026
Multilingual voice AI for Indian call centers: a build checklist
A voice AI line in Indian languages succeeds when it resolves the calls that make up most of the volume, hands the rest to a person with context, and logs every call in the core systems. A build checklist, with results from a line we built that answers in 11 languages.
Tell us which system the AI has to run inside
If that system is not ready for AI yet, tell us that too. An engineer will reply and say plainly whether we are the right firm for it.