Frequently asked questions about Artificial Intelligence consulting
Real questions from prospective and current clients. If yours is not listed here, send us a message and we will answer personally.
Getting started
A proof-of-concept usually takes four to six weeks. That includes data exploration, model training, evaluation and a live demo. Moving from prototype to production adds another two to four weeks, depending on how complex your deployment environment is. If you need the model integrated into a legacy ERP system, for instance, expect the upper end of that range. Simple API deployments on cloud platforms are faster.
At minimum, a structured dataset relevant to the problem you want to solve. This could be a CSV export from your CRM, direct access to a SQL database, or an API feed from your operational systems. We assess data quality during the discovery phase. If your data is messy or incomplete, we will tell you what needs cleaning and whether the project is still viable. In some cases we have helped clients set up basic data collection pipelines before the modelling work begins.
Proof-of-concept projects start around £5,000. Full production deployments with ongoing support typically range from £15,000 to £60,000 depending on the number of models, data sources and integration points. We quote fixed prices for defined scopes, so there are no surprise invoices. Strategy audits sit between £3,000 and £8,000. We publish these ranges because vague pricing wastes everyone's time.
Both. About 40% of our clients have fewer than fifty employees. The deciding factor is not company size but whether you have a clear problem and enough data to train a model. A ten-person e-commerce shop with three years of order history can benefit from demand forecasting just as much as a large retailer.
Technical questions
We deploy on AWS, Google Cloud and Microsoft Azure. If you already have an account on one of these, we use it. If not, we help you set one up. We avoid proprietary platforms that lock you in. Every model we deliver can be exported and re-deployed elsewhere if you decide to move.
Yes. Once the project is paid for, all source code, trained model weights and documentation belong to you. We retain no proprietary rights. You can hand the system to an internal team or another vendor at any point. We think this is the only fair arrangement.
We sign a mutual NDA and, where required, a data processing agreement before any data is shared. All data transfer happens over encrypted channels. We work on your infrastructure whenever possible, so the data never leaves your environment. For projects that require us to hold data temporarily, we use UK-based encrypted storage and delete everything within 30 days of project completion.
This is called data drift, and it is normal. Customer behaviour changes, product lines shift, market conditions evolve. Our support plans include monthly monitoring that flags accuracy degradation automatically. When drift exceeds a threshold we agree on at the start, we retrain the model on recent data. Most retraining cycles complete within a day.
In most cases, yes. We build REST APIs that sit between our model and your application. If your software supports webhooks, scheduled data pulls or direct API calls, integration is straightforward. We have connected models to Salesforce, SAP, Shopify, custom Django apps and several bespoke warehouse management systems. During discovery we map out the integration points and flag any potential friction early.
Working with us
Monthly model performance reviews, automated drift detection, retraining when needed, and up to four hours of ad-hoc consulting per month. Support plans run for a minimum of twelve months and renew quarterly after that. If you have an internal data team that wants to take over maintenance, we provide a handover package with full documentation and a training session.
Yes. Our contracts include a break clause at the end of each phase. If the proof-of-concept does not meet the agreed accuracy targets, you can walk away with no further obligation. You keep all work completed up to that point.
We do. After deployment, we run a half-day workshop covering how the model works, how to interpret its outputs, and how to spot problems. For teams that want deeper knowledge, we offer a two-day technical training covering model architecture, retraining procedures and monitoring setup. Both sessions are priced separately from the project itself.
Our consulting team during a model review session.