Introducing Stoker - AI that reads tenders so you don’t have toIntroducing Stoker - AI that reads tenders so you don’t have toIntroducing Stoker - AI that reads tenders so you don’t have to

Our very first introduction to a prospective new project comes through an invitation to tender. This could be the very start of an exciting new journey for us, as well as for our clients, therefore this step is quite crucial to get right. Our success depends on making smart, fast, and consistent decisions to realise if it's prudent to bid or not.

Teams must navigate a sea of documentation, looking for the critical details that signal a perfect match, all while operating under tight deadlines. But now, thanks to Nearform's AI-enabled agentic tool Stoker, this initial step has become a lot more effective and our decision making process highly optimised.

The tender process, especially in the Public sector, can be notoriously complex. With every opportunity comes a mountain of paperwork, compliance forms, and appendices, not to mention the complex technicalities of the specifications themselves.

There is always the chance of human error and interesting opportunities may be missed in the daily deluge, while time is wasted analysing opportunities that are ultimately not the best fit. Besides, it's important to fully understand what’s being asked, however this could take a team of skilled engineers and senior sales professionals days to go through all the details and extract the basics. Manually sifting through hundreds of pages to assess eligibility, technical fit, and strategic value is a monumental task. It’s time consuming, prone to human error, and can lead to inconsistencies in evaluations across different opportunities.

In order to address this challenge, we ran an internal study to assess the feasibility of automating parts of this complex and time consuming process via AI. The resulting asset is Stoker, aptly named like all Nearform’s AI-enabled software, after Bram Stoker.

Stoker can now extract the necessary details and build the right context enabling us to assess early on whether we’re a good match for the client’s needs. Our goal was to create a tool that could autonomously find relevant opportunities, perform a deep and structured analysis, and deliver a clear recommendation directly to our team. Once this initial insight is available, our team can step in with the human touch and dive deeper into the project. Thanks to Stoker, that very first step of understanding the tender has become much easier and quicker, allowing us to connect the right people with the right opportunities much more efficiently.

The first step in building this tool was to bring our existing process under a microscope and uncover the most repetitive as well as time consuming steps. Once we had the workflow analysed we then applied our AI expertise to build the tool. The team got together and brainstormed on our approach and we saw value in building a multi-agent AI system, where a variety of AI agents could collaborate to tackle different parts of the process right from the discovery to the analysis.

We built Stoker from the ground up using Google Agent Development Kit in Python, tapping into Google’s cutting-edge AI models like Gemini 2.5 pro for in-depth summaries and Gemini 2.5 flash for speedy document analysis. To make sure the AI’s insights matched Nearform’s strategy and technical strengths, we crafted a set of smart prompts that really capture what matters to our business. Everything runs smoothly, cheaply and and on schedule, thanks to automation with GitHub Actions, so the whole process is seamless and hands-off.

Stoker’s automated workflow is simple yet powerful. Here are some of the highlights based on a case study of Stoker’s interaction with Irish government’s eTenders portal. Every day, Stoker’s automated search feature scans the portal for new IT-related tenders. It retrieves all associated documents for each new tender to start its in depth analysis. Stoker has the ability to convert the material files to pdf format if necessary, in order to keep the process and information consistent. Once all the basic material is ingested, the magic begins and a team of AI agents gets to work.

The rule engine handles the analysis behind the scenes, and once it is completed, the system pulls everything together into a set of outputs. These results are delivered in a variety of formats making it further easier to consume. An Excel sheet is produced with relevant data alongside a polished PDF summary for quick reference, as well as a short headline text recommendation for immediate clarity and attention. This information is then automatically broadcast on a dedicated Slack channel, so the team immediately knows of any new opportunities that have been assessed and made ready for further review.

The real power of Stoker is how it turns lengthy, complicated textual information into clear and practical guidance. For every key area we assess the system gives a simple head start for the team, along with a confidence score and an explanation backed by direct references to the original documents.

This means that our commercial team can move from discovery to a confident decision in a much shorter time scale. Stoker cuts through the noise and highlights what really matters, so our experts can focus their time where it counts most, that is building a well-informed and winning proposal for the high-potential opportunities ahead of us.

Stoker has been a recent addition to our team but we are already benefiting from it. In our fast paced, highly competitive environment, where our teams are trying to achieve high impact results, it has been proving a real strategic advantage. This only goes to show Nearform's expertise and keen interest in using Artificial Intelligence in the real world, turning everyday hurdles into a competitive edge.