About us
Intelligent AI Paths started in 2021 when two machine-learning engineers got tired of watching businesses pay six-figure sums for AI projects that never left the prototype stage. We decided to do things differently: smaller teams, faster delivery, and models that run in production from week one.
Our office sits on Mill Lane in Effertz-over-Grady. The team has grown to nine people, including three data scientists, two MLOps engineers, a UX designer, and a project manager. We keep the group small on purpose. Every client talks directly to the person building their system.
Most of our work falls into four areas: natural language processing for legal and medical documents, demand-forecasting models for retail supply chains, computer vision for quality inspection on factory lines, and conversational AI for customer support. We have shipped 47 production systems so far. Thirty-nine of those are still running and maintained by us under ongoing contracts.
We use Python, PyTorch, and a handful of cloud services. Nothing exotic. The value is in how the model fits your data and your workflow, not in the framework.
Our services
Each project starts with a two-week discovery phase. We look at your data, define measurable success criteria, and give you a fixed-price quote before any model training begins.
Natural language processing
We extract structured data from contracts, invoices, medical records, and compliance documents. Our extraction pipelines typically reach 94% accuracy on messy, scanned PDFs after fine-tuning on 500 labelled examples from your own files. Turnaround for a first working prototype is three weeks.
Computer vision
Defect detection on production lines, vehicle counting from CCTV feeds, shelf-stock monitoring in retail. We train object-detection and segmentation models on your images, deploy them to edge devices or cloud endpoints, and monitor drift weekly. One manufacturing client reduced manual inspection time by 62% in the first quarter.
Predictive analytics
Demand forecasting, churn prediction, price optimisation. We connect to your existing databases, build time-series or gradient-boosted models, and deliver predictions through a REST API or a simple dashboard. A grocery chain we work with cut food waste by 19% using daily demand forecasts generated by our system.
Conversational AI
Custom chatbots and voice assistants grounded in your knowledge base. Unlike generic chatbot builders, we fine-tune language models on your documentation so the bot answers questions about your products, policies, and pricing with real accuracy. Setup takes four to six weeks, including integration with your CRM or helpdesk.
MLOps and model maintenance
Already have a model but it keeps breaking? We set up monitoring for data drift, automate retraining pipelines, and handle versioning so your predictions stay reliable month after month. We charge a flat monthly fee, not per-incident.
AI strategy consulting
Not sure where AI fits in your operations? We run a one-day workshop with your team, audit your data assets, and produce a prioritised roadmap with cost estimates. The workshop fee is £1,200 and is credited against any project you commission within 90 days.
How a project works
- Discovery call — 30-minute video call. You describe the problem; we ask about your data, systems, and timeline. No charge.
- Data audit — We review a sample of your data (under NDA) and confirm whether the project is feasible. This takes two to five business days.
- Proposal — Fixed-price quote, scope document, and a timeline broken into weekly milestones. You approve before any work starts.
- Build and iterate — Weekly demos so you can see progress and give feedback. We adjust the model based on your input, not just accuracy metrics.
- Deployment — We deploy to your infrastructure or ours. Load testing, monitoring dashboards, and documentation are included.
- Ongoing support — Optional monthly retainer covers retraining, drift alerts, and priority bug fixes.
Frequently asked questions
How much does a typical project cost?
Small projects like a document-extraction pipeline start around £8,000. Mid-size forecasting or computer-vision systems usually fall between £15,000 and £35,000. Enterprise engagements with multiple models and integrations can go higher. We always provide a fixed quote before starting.
Do I need a lot of data to get started?
It depends on the task. For text classification, a few hundred labelled examples often suffice. Image models need more, typically 1,000 to 5,000 annotated images for good results. If you have less, we can use transfer learning or synthetic data augmentation to close the gap. During the data audit we tell you honestly whether your dataset is large enough.
What industries do you work with?
Retail, logistics, healthcare, legal, and manufacturing make up about 80% of our client base. We have also done projects for energy companies and local councils. The common thread is structured or semi-structured data that humans currently process by hand.
Will my data be secure?
Yes. We sign an NDA before seeing any data. All data stays within UK-based servers unless you specify otherwise. We delete project data 30 days after handover unless you contract us for ongoing maintenance.
How long until I see results?
A working prototype is usually ready in two to four weeks. Production deployment adds another one to three weeks depending on integration complexity. The entire cycle from first call to live system rarely exceeds ten weeks.
Can you work with our existing tech stack?
We integrate with AWS, Azure, and GCP. On-premise deployments are possible too. If you use a specific CRM, ERP, or data warehouse, tell us during the discovery call and we will confirm compatibility before quoting.
Get in touch
Address:
829 Mill Lane, Effertz-over-Grady, Scotland, ZH95 3XI, United Kingdom
Phone:
+44 346 829 5272
Email:
[email protected]
Office hours: Monday to Friday, 9:00 to 17:30 GMT. We respond to emails within one business day.
For urgent production issues, maintenance-contract clients have access to a dedicated out-of-hours number.