Artificial Intelligence FAQ

Honest answers to the questions prospective clients ask us most often. If yours is not here, ring us or send an email.

Common questions

Click any question to expand the answer.

Most projects fall between £15,000 and £80,000. The biggest cost drivers are data preparation (how clean is your data today?), the number of integrations with existing systems, and whether you need ongoing model monitoring.

A proof-of-concept on a single use case typically costs £12,000 to £20,000 and takes four weeks. That gives you a working prototype and enough evidence to decide whether a full build is worthwhile. We quote fixed prices for the proof-of-concept stage so there are no surprises.

The proof-of-concept phase runs about four weeks. After that, production deployment usually takes six to twelve weeks, depending on how many systems the model needs to talk to. A chatbot that connects to a single knowledge base is faster than a demand forecasting system that pulls data from an ERP, a weather API and a calendar service.

We set milestones at the start of every engagement and share progress weekly. If something is going to take longer than planned, you hear about it the moment we know, not at the end.

It depends on the task. Document classification can work well with as few as 500 labelled examples per category. Demand forecasting usually needs at least 18 months of daily transaction data to capture seasonal patterns.

During the data audit we assess what you have and tell you honestly whether it is enough. If it is not, we sometimes recommend a short data-collection phase before starting model development. That is cheaper than building a model on thin data and getting poor results.

Not automatically. Models can drift as the real world changes. A pricing model trained on 2023 data may become less accurate in 2025 because customer behaviour shifts, competitors change their offers, or new products enter the market.

Every deployment we build includes monitoring dashboards that track accuracy metrics in real time. When performance drops below a threshold we agree with you, we retrain the model on fresh data. During the first six months of support, retraining is included at no extra charge. After that, we offer annual maintenance contracts.

We sign a mutual NDA before you share any data. All data is transferred via encrypted channels and stored on UK-based servers. We never use client data to train models for other clients.

When the project ends, we delete all copies of your data from our systems within 30 days unless you ask us to retain it for ongoing support. We are happy to work within your own cloud environment if you prefer that your data never leaves your infrastructure at all. Several of our financial-services clients have chosen this approach.

Yes. We have deployed models into environments running AWS, Azure and Google Cloud, as well as on-premise servers. On the application side, we have integrated with Salesforce, SAP, Microsoft Dynamics, Shopify and various bespoke systems via REST APIs.

During the discovery call we map out your current stack so there are no surprises at integration time. If a particular connector does not exist, we build it as part of the project scope.

That is exactly why we start with a proof-of-concept. At the end of those four weeks, we present accuracy metrics and sample outputs. If the results do not meet the agreed threshold, you can walk away having spent only the proof-of-concept fee.

We set realistic expectations from the start. If we believe a use case is unlikely to succeed with your available data, we say so during the data audit rather than taking your money and hoping for the best. About one in five initial proposals we receive includes at least one use case we advise against, and clients appreciate that honesty.

Every production deployment includes a half-day training session for the people who will use the system daily. We cover how to interpret the model's outputs, when to override its recommendations, and how to escalate issues.

For teams that want deeper technical knowledge, we offer a separate two-day workshop covering model retraining, monitoring dashboards and basic troubleshooting. This is aimed at data analysts or developers who will maintain the system after our support period ends.

Things worth knowing before you call

Industries we have worked in

Logistics and freight, grocery retail, financial services, veterinary healthcare, property management, legal, recruitment, manufacturing, e-commerce, hospitality, insurance, energy, education and media. Each industry has its own data quirks. A logistics company's shipment records look nothing like a law firm's case files, and the models we build reflect that.

If your industry is not on this list, that does not mean we cannot help. It just means we have not listed every sector. Ring us and describe your problem; we will tell you quickly whether it is a good fit.

Our team

We are a group of nine, based in London near Old Street. The team includes machine-learning engineers, a data engineer, a project manager and a commercial lead. Everyone on the technical side holds at least a master's degree in a quantitative field, and most have worked in industry for five years or more before joining us.

You will always have a named project lead who is your single point of contact. We do not hand you off to a junior after the sales call. The person who scopes the project is the same person who oversees delivery.

Still have questions?

We would rather spend 30 minutes on the phone with you than have you guess. The consultation is free and there is no sales pitch at the end.

Phone: +44 20 6699 9504
Email: [email protected]

Send us a message