Artificial Intelligence FAQ

A Starter-tier project runs about six weeks from kick-off to deployment. Professional projects take eight to twelve weeks. Enterprise engagements vary, but most deliver a first production model within ten weeks. The data-audit sprint at the beginning is what determines the real timeline: if your data is clean and well-labelled, we move fast. If it needs significant preparation, we add a week or two to the audit phase and adjust the schedule before any modelling begins.

Not necessarily. Every engagement begins with a paid data-audit sprint where we review your schemas, identify gaps and clean what needs cleaning. If your data is already well-structured, that sprint finishes faster and we move on to modelling sooner. What helps us most is access: database credentials, API keys, sample exports. The less time we spend chasing permissions, the more time goes into actual analysis.

You do. All intellectual property transfers to you upon full payment. We retain no licence to use your data or models after the project ends, unless you explicitly ask us to host or maintain the system. We will hand over all source code, training scripts, configuration files and documentation in a private Git repository you control.

We work on UK-based cloud infrastructure by default (AWS eu-west-2 or Azure UK South). If you need data to stay on-premise, we can set up a secure development environment on your own servers. All data in transit is encrypted with TLS 1.3, and data at rest uses AES-256 encryption. We delete all copies of your data from our systems within 14 days of project completion, and we provide written confirmation when we do.

We set a target metric (accuracy, F1, mean absolute error, or whatever makes sense for your use case) during the proposal phase. If we cannot hit it after the agreed number of iteration cycles, we will explain why, share what we learned about your data and recommend next steps. You will not be charged for additional cycles beyond the agreed scope. In practice this happens rarely because the data audit catches most blockers before modelling begins.

Yes. We have built connectors for Salesforce, HubSpot, SAP, Xero, Shopify, PostgreSQL, MySQL, MongoDB, BigQuery and several bespoke internal systems. If your tool has a REST or GraphQL API, we can connect to it. The Professional package includes one integration; the Enterprise package covers multiple systems. Extra integrations can be added for £1,200 each.

Every response gets a confidence score. When the score falls below a threshold you set (we recommend 0.65 as a starting point), the bot tells the user it is not sure and offers to connect them with a human agent. The conversation transcript transfers automatically, so the agent does not ask the customer to repeat themselves. You can review low-confidence conversations in the management dashboard and add new source material to fill the gaps.

Every package includes a support period: 30 days for Starter, 90 days for Professional, 12 months for Enterprise. During that period we fix bugs, answer questions and monitor model performance at no extra cost. After the included period ends, you can sign up for our extended-support add-on at £600 per month, which covers monitoring, bug fixes and a four-business-hour response time.

Bank transfer (BACS or Faster Payments) is our default. We can also accept payment via Stripe if you prefer to pay by card. Fixed-scope projects are invoiced 50% at kick-off and 50% on delivery. Retainers are billed monthly in advance. If you need a different payment schedule, mention it during the discovery call and we will do our best to accommodate it.

Starter projects are handled by a single ML engineer who also manages the data pipeline work. Professional engagements typically involve two people: one ML engineer and one data engineer. Enterprise projects get a dedicated project lead plus two or three engineers depending on scope. You will always have a single point of contact regardless of team size.