Document automation
Local authorities receive thousands of documents every week: planning applications, licensing renewals, complaints, subject-access requests. Staff spend hours reading, categorising and forwarding them.
Our document automation pipeline uses natural-language processing to read each incoming file (PDF, Word or scanned image via OCR), classify it against your existing taxonomy and assign it to the correct team. The model learns from corrections your staff make, so accuracy improves over time without any retraining on our side.
What you get
- A classification model trained on your own document history (minimum 500 labelled examples needed).
- An integration layer that connects to your email inbox, SharePoint folder or case-management system.
- A review dashboard where staff can confirm or override the model's decision in one click.
- Monthly accuracy reports emailed to your project lead.
Typical delivery: six weeks from kick-off to production for a single document type. Each additional type adds roughly two weeks.
Demand forecasting
Knowing how many residents will phone your helpline next Monday, or how many tonnes of recycling a depot will handle in March, lets you roster staff and order supplies without guessing. Our forecasting models use your historical volumes plus external signals (weather, school terms, public holidays) to produce daily or weekly predictions.
We have built forecasting systems for waste-collection routes, housing-repair scheduling and library footfall. In each case the client had at least 18 months of timestamped records, which is the minimum we need to capture seasonal patterns.
What you get
- A time-series model (Prophet or LightGBM depending on data shape) retrained automatically each week.
- Forecasts pushed to a simple web page or exported as CSV for import into your rostering tool.
- Confidence intervals so managers can see best-case and worst-case alongside the point estimate.
- A quarterly review call where we assess drift and adjust features if needed.
Typical delivery: four weeks for a single metric. Bundles of three to five related metrics take eight weeks.
Resident-facing chatbots
A well-built chatbot can answer the twenty questions that account for 60 per cent of your contact-centre calls: bin-collection days, council-tax payment dates, opening hours, parking-permit applications. Ours run on a retrieval-augmented generation (RAG) architecture, which means the bot pulls answers from your own published content rather than making things up.
We configure the bot to refuse questions outside its knowledge base and to hand the conversation to a human agent when it cannot help. Every conversation is logged so you can spot gaps in your public-facing information.
What you get
- A chatbot widget that embeds on your website and optionally on WhatsApp or Facebook Messenger.
- An admin panel where your comms team can add, edit or remove knowledge articles without touching code.
- Conversation analytics: most-asked topics, resolution rate, handoff rate.
- Hosting on UK-based servers; no resident data leaves the jurisdiction.
Typical delivery: five weeks including content migration and user testing.
How every engagement works
Regardless of which service you choose, the project follows four stages. We do not move to the next stage until you sign off the previous one.
Discovery
We visit your offices (or join a video call), review the data you have, and write a one-page brief that defines the problem, the success metric and the timeline. This stage is free.
Proof of concept
Using a sample of your data we build a working prototype and test it against the agreed metric. You see real results on real data before committing to a full build. This stage typically costs between £2,000 and £5,000.
Production build
We harden the prototype, connect it to your live systems, and train your staff to use and maintain it. Costs depend on complexity; we quote a fixed price after the proof of concept.
Support
For twelve months after go-live we monitor model performance, fix bugs and answer questions at no extra charge. After that, ongoing support is available on a monthly retainer.
Indicative pricing
These ranges cover the full engagement from discovery to go-live. VAT is excluded. We always confirm a fixed price before any paid work starts.
| Service | Typical range | Timeline |
|---|---|---|
| Document automation (single type) | £8,000 – £15,000 | 6 weeks |
| Demand forecasting (single metric) | £5,000 – £10,000 | 4 weeks |
| Demand forecasting (3–5 metrics) | £12,000 – £22,000 | 8 weeks |
| Resident chatbot | £10,000 – £18,000 | 5 weeks |
| Data dashboard | £4,000 – £9,000 | 3 weeks |
| Staff AI training workshop (full day) | £1,200 per session | 1 day |