Beyond the Hype: Helping Businesses Turn AI Experimentation into Real Impact

By Matt Carroll, CEO, and Nigel Garner, CTO, Nimble Approach

AI – and generative AI more specifically – has dominated headlines for nearly two years. Yet for many organisations it remains more promise than progress. Why?

Because most AI activity sits disconnected from the day-to-day work that actually drives value.

At Nimble Approach, we believe it doesn’t have to be this way.

AI will transform the way businesses grow, operate, and innovate — but only if it’s approached as a business transformation challenge, not just a technology one.

Where AI Should Be Making a Difference

Organisations pursuing AI are ultimately chasing three outcomes:

According to IBM research, executives expect 62% of AI investment to be directed toward products, services, and business-model innovation by 2030.

Further, similar to how digital-first or digital-natives leapfrogged some of the legacy incumbents, it’s anticipated that AI-first (rather than AI-bolt ons) will see 70% greater improvements in productivity, highlighting the need for AI to be part of the strategy.

AI can support both. Traditional AI has already proven its value in areas like forecasting, customer service, and operational efficiency. Generative AI goes further, enabling conversational, contextual, and creative capabilities that fundamentally change how work gets done.

The missing link isn’t potential. It’s execution.

What’s Holding Businesses Back?

We see three shifts every organisation has to tackle before AI value becomes real:

1. Modernisation: The Prerequisite for Intelligence

Outdated tech and data platforms do more than just slow you down; they actively block the cognitive leaps AI promises. Legacy systems simply were not built to handle the throughput, agility, or iterative learning that Generative AI demands.

The biggest friction point we see is that most legacy environments are opaque to AI.

• The Data Disconnect: AI requires clean, real-time data pipelines to function. Legacy systems often lock data away in fragmented on-premise silos or “brittle” delivery pipelines that cannot support AI at scale.
• The “Agentic” Barrier: We are moving rapidly from Chatbots (that talk) to Agentic AI (that act). For an AI agent to autonomously execute tasks, it needs to interface with your systems via clear, standard protocols. Monolithic legacy applications generally lack the API-first architecture required for agents to connect and perform work safely and securely.
• The Decision: Patch or Platform? These constraints force a critical choice: attempt to bolt AI onto aging infrastructure, or invest in an AI-native foundation. The latter means building modern application and data platforms – often in cloud or hybrid environments – designed for scalability, security, and continuous learning. This doesn’t require a multi-year “Big Bang” rewrite; in fact, we strongly advise against it.

At Nimble Approach, we help clients modernise to enable AI. Using the Strangler Fig pattern, we build modern, cloud-native services around your legacy core to deliver value in months, not years.

2. AI Governance: From “Red Tape” to Competitive Advantage

As AI governance and regulation gathers pace, most notably with the EU AI Act and the UK’s published AI Cyber Security Code of Practice – governance has shifted from optional to essential. But treating it purely as a compliance exercise is a strategic mistake.

In 2026, AI governance is a competitive enabler. Just as high-performance brakes allow an F1 car to drive faster, robust guardrails allow organisations to deploy AI with greater confidence, speed, and scale.

The Cost of Getting It Wrong

The risks of “moving fast and breaking things” are no longer theoretical.

• Financial impact: Research from EY shows that 99% of the 975 global organisations that responded to its survey have already experienced financial loss linked to AI-related risks, with the average cost of failure reaching $4.4 million.
• Reputational and operational exposure: Failures increasingly erode customer trust, attract regulatory scrutiny, and stall AI programmes.

The Trust Premium

Governance doesn’t just reduce downside – it creates value.

PwC’s 2025 analysis found that organisations investing in Responsible AI achieve valuations up to 4% higher than peers, driven by stronger stakeholder trust and faster, more sustainable innovation cycles.

The Nimble Approach: Security by Design

We believe that if security slows you down, it’s designed wrong. We help clients move away from manual “tick-box” compliance toward automated governance that lives within the code itself.

• Secure by Design: We align with the National Cyber Security Centre’s (NCSC) principles, integrating security into the CI/CD pipeline from day one. Crucially, we move beyond simple Role-Based Access Control (RBAC) to Attribute-Based Access Control (ABAC) within data platforms. This ensures models cannot “leak” data, dynamically restricting retrieval based on the user, data sensitivity, and query intent, rather than just a static job title.
• Zero Trust for an Agentic World: As you move toward Agentic AI – where bots actively perform tasks like payments or data retrieval – traditional perimeter security fails. We implement Zero Trust architectures, ensuring that every AI agent has strict, verified limits on what it can access and execute.
• Observability as Standard: You cannot govern what you cannot see. We build “observable” systems with deep telemetry, allowing you to monitor AI behaviour in real-time and detect “drift” or hallucinations before they impact customers.

Taken together, these capabilities point to a simple truth: governance can’t be something you bolt on after the fact. In an agentic, AI-driven world, it has to be built into the foundation. That’s the only approach that scales without slowing teams down.

3. People & Process: The “Project to Product” Gap

The third friction point is organisational muscle memory. Many businesses are trying to deliver probabilistic systems using deterministic delivery models.

Two patterns show up repeatedly:

• The “Feature Factory” Trap: Traditional IT delivery focuses on shipping features on a fixed date. AI doesn’t work that way. AI development is hypothesis-driven; you don’t know if a model can solve a problem until you test it. Organisations that stick to delivering rigid shopping lists of features will struggle to learn genuine insights and iterate fast enough to get the value from AI.
• The Product Management Void: Building an AI agent is different from building a web form. It requires Product Managers who:
  • understand data as a strategic asset,
  • can define “guardrails” rather than just “user stories,”,
  • lean into ambiguity and uncertainty, and are comfortable managing the non-deterministic nature of GenAI (where the answer isn’t always the same),
  • can orientate towards clear business outcomes.

What’s required is a deliberate shift in mindset. AI success depends on moving from project thinking — budgets, timelines, and milestones — to product thinking, where value, outcomes, and learning velocity define progress.

How Nimble Approach Closes the Gap

We help clients move from “Project thinking” (budgets, timelines, milestones) to “Product thinking” (value, outcomes, iteration).

• Hypothesis-Driven Delivery: We don’t promise a feature list; we promise to solve a business problem to achieve an outcome. We use the “Thin Slice” approach to prove value early – testing the riskiest assumption first (e.g., “Can the AI actually answer this customer query accurately?”) before building the rest of the infrastructure.
• Empowerment, Not Replacement: The narrative that “AI will replace us” breeds resistance. We flip the script. By involving your teams in the design of the AI tools – showing them how AIDD (AI-Informed Digital Delivery) removes the drudgery from their day – we turn sceptics into champions.

Why Now? The Cost of Waiting has Changed.

Two years ago, “waiting and seeing” was a valid strategy. The technology was volatile, and the use cases were unproven. That window has closed.

We are seeing a definitive shift in the market. The early adopters who spent 2024/25 testing GenAI in isolated “sandboxes” are now operationalising it.

• The Widening Gap: The difference between an AI-enabled organisation and a legacy one is no longer just efficiency; it is capability. Competitors who have modernised their data platforms can now deploy new agentic workflows in weeks. Those still wrestling with on-premise monoliths are finding the gap impossible to close with sheer headcount.
• The Governance Tipping Point: With new governance standards becoming more mature, the “Wild West” era is over. Organisations need partners who can navigate this new landscape not just legally, but technically, baking compliance into the code itself.

Practicing What We Preach

Like most organisations, we know that AI adoption doesn’t start with policy – it starts with behaviour. People are already using AI tools, whether officially sanctioned or not – 90% of respondents to the 2025 DORA report said they were using AI at work.
That’s why we apply AI-Informed Digital Delivery (AIDD) within our own business. We use AI to accelerate how we scope, design, and build software – automating low-value tasks while enabling our teams to focus on architecture, quality, and outcomes.

This hands-on experience matters. We understand the friction points of adoption because we’ve lived them. When we advise clients on modernisation, governance, or delivery, it’s grounded in operational reality and not just theory.

Getting Started

The path to AI impact isn’t a leap of faith; it is a series of focused, high-velocity steps.

At Nimble, we don’t ask you to commit to a multi-year roadmap or a “Big Bang” migration. We identify a single, high-value vertical in your business – be it customer onboarding, claims processing, or internal knowledge retrieval – and we apply our Shorter Time to Value methodology:

Don’t let the scale of the challenge paralyse you. Start small, move fast, and prove the value.

Get Started