Why “Hey Geraldine” Beats Building Your Own Chatbot

Local authorities are under pressure to deliver better digital services while keeping costs under tight control. From a digital delivery perspective, the choice around AI chatbots often boils down to a simple question: should we build our own, or adopt a proven solution like Hey Geraldine?

Why in-house builds get expensive

On paper, a 6–8 week in‑house build can look manageable. In reality, even a focused AI chatbot pilot quickly becomes a significant investment in staff time.

To reach a safe, usable pilot, you typically need:

• Product or delivery management

• Developers with AI and integration experience

• Content and service design

• User research or business analysis

• Support from data protection, security and architecture

  
  

For a typical council digital team, this usually equates to around 60–100 person‑days across those roles to deliver a first working version. Once you apply realistic blended day rates for local‑government digital roles, that translates into a staff‑time equivalent of roughly £38,000–£60,000 for a modest 6–8 week pilot. With additional reliance on specialist AI contractors, the effective cost can rise further, towards £70,000–£90,000 for the same period.

By contrast, Hey Geraldine is a fixed‑price tool at £25,000. In other words, in many scenarios it costs less than a single in‑house build phase, while also avoiding the risk of overruns, rework and false starts. And that comparison is before you account for ongoing support, enhancements and the opportunity cost of taking your internal team away from other priority projects for up to two months.

The hidden cost: knowledge base construction

The visible chatbot interface is only half of the challenge. The real work – and risk – sits in designing and maintaining a robust knowledge base.

This includes:

• Identifying and prioritising the right source material

• Cleansing, structuring and tagging content for AI use

• Resolving contradictions and reflecting local policy nuance

• Designing governance so updates flow through safely and consistently

• 
  

Done entirely in‑house, this absorbs time from service leads, information governance, and web/comms teams. Each council effectively pays to rediscover the same patterns: how to structure FAQs, how to represent complex journeys (like adult social care eligibility), and how to make sure an AI assistant is both helpful and safe.

Without prior patterns to follow, teams can easily spend weeks on trial and error before the chatbot feels trustworthy enough to expose to residents or staff.

Why reusing “Hey Geraldine”-type solutions is cheaper

Adopting an existing, proven solution like Hey Geraldine means buying into a working platform and a set of tried‑and‑tested patterns, not just a piece of software.

Core problems have already been solved:

• LLM orchestration and guardrails

• Conversation flows and escalation routes

• Monitoring, analytics and feedback loops

• Safety mechanisms for sensitive domains like adult social care

  
  

This allows your digital team to focus on configuration, change management and adoption, rather than inventing a chatbot stack from first principles. You largely avoid the initial 60–100 person‑day build period, and significantly reduce the risk that your first iteration fails to meet resident or member expectations.

How Datnexa lowers delivery risk and cost

Datnexa helps councils go further by knowing the path to tread. Having already delivered AI assistants in local government, we understand:

• Which steps add genuine value and which can be streamlined

• The right order to tackle discovery, pilots and scaled roll‑outs

• The common pitfalls in areas like governance, consent and safeguarding

  
  

On the knowledge‑base side, Datnexa brings reusable patterns: content templates, tagging schemes, testing approaches and playbooks for engaging service leads and information governance efficiently. We have experience working alongside PCC colleagues and running cohorts of local authorities, so costs and learning are shared rather than duplicated.

For each council, this turns what could be an open‑ended in‑house build into a shorter, more predictable engagement with clearer milestones, lower staff‑time burn, and faster realised value. In simple terms: you get a safer, better‑tested AI assistant for less money and less pressure on your own digital team.

If you would like to explore what this could look like in your authority, Datnexa can provide a short options paper and indicative timeline tailored to your current digital roadmap.