The people behind your AI software
We started Ai Dynamic Clarity in 2021 because we kept seeing the same pattern: businesses buying expensive AI licences, then struggling to connect them to their actual workflows. The tools sat unused. Budgets were wasted. Teams grew sceptical.
So we took a different approach. Instead of selling a product and walking away, we embed with your operations team for the first few weeks. We learn how your data moves, where it breaks and what a successful outcome looks like in your specific context. Only then do we write a single line of code.
How we got here
A brief timeline of the decisions that shaped our work.
2021: First client, first lesson
Our founding project was an invoice-extraction tool for a logistics company in Glasgow. The model worked well in testing but failed on scanned faxes that arrived from one particular supplier. That taught us to always test with the messiest data first, not the cleanest. We rebuilt the pipeline with a pre-processing step for low-resolution scans, and the system has been running without manual intervention since.
2022: Expanding into customer triage
A mid-sized insurance broker asked us to classify incoming claims emails by urgency. We trained a classifier on 18 months of their ticket history and deployed it in shadow mode for three weeks. After going live, their average response time for high-priority claims dropped from 14 hours to under 90 minutes.
2023: Demand forecasting for retail
We built our first inventory-prediction model for a Scottish food distributor with 1,200 product lines. The model runs nightly, compares predicted demand against current stock levels and sends re-order alerts by 6 a.m. Spoilage costs fell by 22% in the first quarter.
2024 and beyond
We now serve clients in logistics, retail, financial services and healthcare administration. Our team has grown to nine people, and we have moved into a permanent office in High Bode End. The focus remains the same: practical tools, honest performance metrics, no hype.
What guides our work
These are not abstract principles pinned to a wall. They shape how we scope projects, write contracts and handle disagreements.
Measure before you promise
We run a data audit before quoting a project. If the data quality is too low to produce a useful model, we say so upfront rather than billing for months of work that leads nowhere. Two prospective clients last year heard "this is not a good fit for AI" from us. Both appreciated the honesty and came back later with better-prepared datasets.
Transparency in performance
Every deployed model comes with a monitoring dashboard that shows accuracy, latency and error counts in real time. We do not hide behind averages. If the model drifts below the agreed accuracy threshold, our alerting system notifies both your team and ours so we can retrain before it affects your operations.
Fixed-price discovery
The discovery and data-audit phase is always a fixed fee, agreed before we start. You get a written feasibility report at the end, regardless of whether you proceed. No open-ended consulting invoices during the exploration stage.
Your data stays yours
Client data never leaves your infrastructure unless you explicitly authorise it. We can train models on your servers, in your cloud tenancy or in a dedicated environment that we provision and you control. At the end of a project, all data copies on our side are deleted and we provide written confirmation.
The team
Nine people, mostly based in Scotland, with backgrounds in applied maths, software engineering and operations management.
Fiona Rennie
Founder and lead engineer
Fiona spent eight years building data pipelines at a Scottish energy company before starting Ai Dynamic Clarity. She writes most of the initial model prototypes and leads the data-audit phase for new clients. Outside work she runs a weekly coding workshop for secondary-school students in Edinburgh.
Marcus Holt
Operations director
Marcus handles project timelines, client communication and the commercial side of the business. He previously managed supply-chain operations for a UK grocery retailer, which is why he understands the real-world messiness that AI tools need to survive. He is the person you will speak to first when you call us.
Ravi Kapoor
Machine learning engineer
Ravi joined in 2022 after completing a PhD in natural language processing at the University of Glasgow. He built our customer-triage classifier and maintains the retraining pipelines. His speciality is getting high accuracy from small, noisy datasets, which is exactly what most of our clients have.
The rest of the team includes two backend developers, a DevOps engineer, a data analyst and an office manager. We hire slowly and only when project demand justifies it. Every engineer on the team has deployed at least one production model before joining us.
Results from real projects
We share these figures with permission from the clients involved. Ranges reflect variation across different deployment periods.
Since 2021 we have completed 23 projects for 16 clients. Fourteen of those clients are still under active support contracts, which means they continue to use the tools we built. Our average project duration from signed contract to live deployment is nine weeks. The shortest was four weeks for a straightforward document-extraction task; the longest was sixteen weeks for a multi-source demand-forecasting system with six data integrations.
Client-reported cost savings range from 15% to 40% on the specific processes we automated. We track these numbers jointly with the client for the first six months after deployment. If a model underperforms the feasibility estimate by more than ten percentage points, we retrain at no additional charge.
We carry professional indemnity insurance and are registered with the Information Commissioner's Office. All client data handling follows GDPR requirements, and we conduct annual penetration tests on our infrastructure.
