Introduction
While the recent rise of AI technology has opened many new avenues for businesses, business owners now face the challenge of choosing the right AI solutions. But if your AI strategy is limited to chatbots, you’re missing out on the potential of agentic AI: technology designed to perform complex tasks much like humans do.
Data modelling: the vital ingredient
In the world of data consulting, data modelling is a critical activity, capturing the way a business works in a form that data platforms can use. The BEAM* (Business Event Analysis & Modelling) method is the gold standard for data modelling, but it is an advanced methodology that demands time from data experts and business stakeholders. These demands mean that many organisations can’t adopt BEAM* in full – we decided to build an AI agent to overcome this problem.
Chatbots vs Agents: it's all just AI, isn't it?
It’s worth pausing to understand the differences between AI agents and chatbots. While both rely on similar AI technologies, there are key distinctions:
- Autonomy Agents can plan and make decisions independently to achieve a goal, whereas chatbots respond to user prompts. A chatbot can draft a project update , but an agent can monitor a project, decide which tasks must be completed, prepare a report, and ask for missing information.
- Goal-orientation Agents can perform complex, multi-step tasks and adapt to the changing context to achieve a goal: for example if asked to ‘prepare a report’, an agent will work out and complete all the steps required, without human intervention. Chatbots typically answer user questions within a conversational flow.
- Operation and maintenance AI agents require more robust setup, monitoring and governance because they can work independently and often have access to corporate systems like email, CRM and ERP systems. By contrast, chatbots are simpler to deploy and manage, operating within the scope of a conversation.
The case for the BEAM* Agent
We use data models to describe the information used by businesses for decision-making, reporting, or workflows. The problem is that capturing this information involves spending time with stakeholders, and requires skilful facilitation to ensure the outputs are of high quality. The complex, multi-step nature of this process, and the need to adapt to the conditions led us to consider whether an AI Agent could do this work.
Faster results - without hallucinations
Our BEAM* Agent is an advanced goal-based AI agent able to turn workshop notes into a wide range of client-ready deliverables, including spreadsheets, presentations, and diagrams. Guided by user input, it works through tasks iteratively – refining and improving outputs to meet client requirements. This enables our project teams to rapidly organise complex source data into the format they need, saving time and increasing consistency.
Crucially, the BEAM* Agent does not create or guess at information beyond these sources. We set it up to work like a human expert would: using source materials and requesting clarification where further detail is needed. Outputs stay grounded in this data meaning they are accurate, transparent and reliable. The agent uses AI to reason about the source information, and standard Python code to create the outputs: this approach virtually eliminates hallucinations.
The power of the human-augmented approach
The BEAM* Agent acts as a skilled assistant for our expert data modellers. We gather source data from the client, and enrich it through workshops with the client. We then put the agent to work: it reviews and processes materials to start building the required outputs and data model. Throughout the process, our experts check and refine the work, giving clients the benefit of their experience combined with the speed of execution from the agent. This human-augmented approach is at the heart of our AI strategy for clients, showing the value of real-world human experience in the era of AI.
The value: speed to results
We’re using this with clients today. For a housing business, the BEAM* Agent used data about letting and repairs to build BEAM* data models. We were quickly able to show the client how they could gain more insights into their data, identifying gaps, weaknesses, and strengths. This means faster decisions and earlier benefits, giving clients the power of BEAM* more rapidly than has been possible in the past.
