---
title: "The boring AI use cases are the ones that work | SamuylovAI"
description: "On the main stage of the first GenAI London, Denis moderated a reality check on enterprise AI. Where the returns are: defined workflows, governance built into the system, and control over the models."
url: https://www.samuylov.ai/stories/the-boring-ai-use-cases-are-the-ones-that-work
language: en
---

Article · October 24, 2025

# The boring AI use cases are the ones that work

Fraud detection. Developer productivity. KYC and AML checks. Not the demos that fill keynotes, but recurring work with real users and results you can measure. That is where the returns are.

That was the case made on the main stage of the first [GenAI London](https://www.genailondon.uk/), at the QEII Centre on 24 October 2025\. Denis moderated _GenAI in 2025: Time for a Reality Check_, with Adrian Lepers ([Hugging Face](https://huggingface.co/)), Dr. Petar Tsankov ([LatticeFlow AI](https://latticeflow.ai/)), Alexander Taboriskiy ([Mentiora AI](https://mentiora.ai/)) and Anne-Marie Buzatu ([ICT4Peace Foundation](https://ict4peace.org/)).

![Panellists beneath the GenAI reality-check session title.](https://www.samuylov.ai/assets/4954b537/stories/the-boring-ai-use-cases-are-the-ones-that-work/panel.webp)

Main stage, QEII Centre, London, 24 October 2025.

## Start with an outcome, not a pilot

The backdrop was MIT NANDA’s report [_The GenAI Divide_](https://mlq.ai/media/quarterly%5Fdecks/v0.1%5FState%5Fof%5FAI%5Fin%5FBusiness%5F2025%5FReport.pdf) and its 95% headline. The panel’s reading: that failure rate was predictable, given how many organisations had rushed out pilots with little strategy behind them. The returns come from starting with a business outcome and picking the boring use case that serves it.

## Governance you can execute

Governance only works when it is built into the technology and can be executed. A policy written up and hung on the wall governs nothing. Where an error can cause harm, that means safeguards inside the system and clear responsibility for the outcome.

## Control your stack

Open models are no longer the alternative; they are the baseline. Smaller domain-specific models, sparse mixture-of-experts architectures and quantisation make them worth weighing against the task, the running cost and the control you need.

## What it means for your project

Our own case studies are boring in exactly this sense: [classifying documents](https://www.samuylov.ai/stories/document-classification-for-a-global-procurement-provider), [enriching lead data](https://www.samuylov.ai/stories/company-data-enrichment-for-an-office-solutions-provider), [changing airline tickets](https://www.samuylov.ai/stories/ticket-change-copilot-for-a-large-travel-group). Recurring work, a clear owner, results you can check.

Pick a recurring task with a known baseline, an owner and a cost of error. That gives a pilot a clear test before rollout. See [advisory & implementation](https://www.samuylov.ai/consulting), or [Start a conversation](https://www.samuylov.ai/contact).

[Download as PDF](https://www.samuylov.ai/stories/the-boring-ai-use-cases-are-the-ones-that-work/pdf)

## Have a project in mind?

Tell us what you are building or improving. We'll help you choose a practical next step.

[Book an intro call](https://www.samuylov.ai/contact?book=1&topic=other)

## More stories

[All stories](https://www.samuylov.ai/stories)

[Lead-data enrichment for the sales team of a workplace coffee and water supplier](https://www.samuylov.ai/stories/company-data-enrichment-for-an-office-solutions-provider) – Case study · B2B Services · November 2025 · To a working proof of value: 3 days · In production: 11 months

[The GenAI bottleneck was never the model](https://www.samuylov.ai/stories/the-genai-bottleneck-was-never-the-model) – Article · September 2025
