AI implementation isn't the hard part. Neither is integration. It's knowing what business you're trying to build.
By Gabbi Stubbs
There's no shortage of advice about AI and why AI initiatives fail to move beyond pilots. The usual suspects? Poor data. Disconnected systems. Legacy technology. Lack of integration. Insufficient skills. And, of course, the answer is often more technology.
There's truth in all of this. But we're missing a much larger point.
We're in danger of approaching AI as a technology problem when it is fundamentally a business transformation challenge.
That distinction matters. Because if we get the starting considerations wrong, we risk spending millions integrating AI into businesses that haven't stopped to question whether their existing processes, operating models or technology architectures are fit for the future.
The problem with the current narrative
Much of the current conversation focuses on moving AI from experimentation into production. The argument goes something like this: building a proof of concept is relatively easy; integrating AI into enterprise data, applications and workflows is the hard part.
Fair. Integration is important. Context matters (immensely). Governance matters (more than anything). And a successful pilot is a long way from a reliable, scalable business capability.
But integration is not a strategy. Nor is it the outcome.
You can integrate AI perfectly and still fail to create meaningful business value.
You can automate a process nobody needs. Optimise a metric that matters to no-one. Build an impressive agent that adds cost rather than value. Or make an existing, inefficient process run faster without questioning why it exists in the first place.
The challenge isn't simply making AI work within the business as it operates today. It's understanding how the business should operate tomorrow, and determining where AI can fundamentally change what's possible.
Follow the money. And question who's selling the solution.
One of the issues with the current AI conversation is the out-weighted influence of AI platforms and technology vendors.
Vendors are rightly talking about data foundations, integration layers, orchestration, agents, platforms and the infrastructure required to make AI work. These are products they sell. They have a commercial interest in positioning their existing platforms, expanding their addressable markets and making their technology central to the AI opportunity.
That doesn't make their advice wrong. But it does mean we need to understand the lens through which it is being presented.
A vendor selling a customer data platform will naturally emphasise unified customer data. A cloud provider will talk about infrastructure and model deployment. A SaaS provider will position AI as the next layer of capability within its application. An agent platform will tell you that orchestration is the missing piece.
Each may be right about part, even large parts, of the problem. Few necessarily have the complete picture.
And when the people shaping the conversation are predominantly technology vendors, it's hardly surprising that the proposed solutions tend to be technology-centric.
The result? Businesses risk starting with the technology they can buy rather than the business outcomes they need to achieve.
They build around existing architectures, extend existing platforms and add another layer of complexity. Meanwhile, the more fundamental questions about business models, customer value, organisational design and how work should actually get done receive far less attention.
AI strategy cannot simply become another route to selling more software.
Business leaders and Boards need to be able to distinguish between a genuine business requirement and a vendor's preferred architecture. They need to understand what is essential, what is optional, what can be simplified and what might no longer be necessary at all.
Sometimes the answer will be more technology. Sometimes it will be better integration. And sometimes the best answer will be to remove complexity, redesign a process or not use AI at all.
Start with the business, not the technology
So, what should organisations be considering when developing an AI strategy or moving into implementation? I would start with six interconnected questions.
What business problem are we actually trying to solve?
Not what can we do with AI. Not which use cases look impressive in a demonstration. And not where can we deploy the latest agent. What are we trying to achieve? Are we trying to improve customer experience, increase revenue, reduce cost, accelerate decision-making, improve risk management or create an entirely new product or service? What is the current baseline? What does success look like? What is the potential value? And how will we measure it? This sounds obvious, yet too many initiatives begin with a technology capability and work backwards to find a problem it might solve. AI should be assessed against the same commercial discipline as any other significant business investment. If conventional automation, a process change or a simpler software solution delivers a better outcome at lower cost and risk, why wouldn't we choose it?
Do we need to integrate the existing architecture, or rethink it?
This is where the integration conversation becomes interesting. Much of the current thinking assumes that the answer is to connect AI to existing data, applications and workflows. For many businesses, that is entirely sensible. They have substantial investments in technology that works, and extending those capabilities is often the most practical route forward. But we shouldn't assume that every existing system, process or architectural decision deserves to be preserved. AI creates an opportunity to rethink how work is organised, how decisions are made and how applications interact. It may enable simpler workflows, more composable architectures and new ways of delivering customer and employee experiences. The answer won't be the same for every organisation. But it should be a conscious strategic choice, not a default dictated by a vendor's product roadmap.
Are we fixing the underlying business, or simply automating its existing inefficiencies?
This is one of the biggest opportunities — and the biggest risk. If a business has fragmented accountability, poor decision-making, unnecessary hand-offs or processes designed around organisational silos, introducing AI will not fix those problems. In some cases, it will amplify them. Before automating a workflow, understand why it works the way it does. Identify unnecessary steps, duplicated effort, conflicting incentives and decisions that could be made closer to the customer or the point of action. Then redesign the process. The goal shouldn't be to make today's processes faster simply because we can. It should be to determine what the process needs to achieve, how it could operate differently and where AI creates a genuine advantage.
What data, context and controls does the use case actually need?
Yes, data matters. Immensely. AI needs access to relevant information, context and business rules to make useful decisions and take appropriate action. Data quality, identity, permissions, privacy and security are fundamental considerations. But the answer isn't automatically to unify every data source before doing anything useful. Different use cases will have different requirements. Start with the use case. Determine what information is required, where it resides, whether it is trustworthy and importantly, what the AI is authorised to do with it. Then build the necessary foundation. The objective is not the most sophisticated architecture. It's the simplest architecture that delivers the required outcome safely and consistently.
Who is accountable when AI becomes part of the business?
This is where the conversation moves well beyond technology. Realistically, technology is no longer the bottleneck; organisational culture and design is. Without up-skilling AND ownership and accountability clarity or people-led governance, AI software will remain largely unused, used in a misguided fashion or used unsafely. These aren't questions that can be delegated to IT, the data team or a vendor. They require business leadership, operational ownership, risk management and, where appropriate, Board oversight. Governance isn't simply a set of policies or an approval process before launch. It needs to operate throughout the lifecycle of the capability. And accountability cannot be outsourced to the model, the platform or the system integrator.
Are we measuring activity, or actual value?
The number of pilots completed, agents deployed, employees trained or workflows automated tells us very little about whether an AI strategy is succeeding. We need to measure what changes as a result. Is revenue increasing incrementally? Are costs genuinely reducing, or simply moving elsewhere? Is customer experience improving? Are decisions more accurate? Is risk being reduced? Are employees spending more time on higher-value work? Is the business creating capabilities or services that weren't previously possible? The test is not whether AI works in a demo. It's whether it delivers sustained, measurable value in production.
The operating model matters as much as the technology
There's another dimension that deserves greater attention: how the organisation itself needs to change.
AI will challenge existing assumptions about roles, responsibilities, decision rights, team structures and the way work flows across a business.
It will also expose organisational problems that technology cannot resolve.
If marketing, sales and service operate to different objectives, better AI won't magically make them align. If accountability is fragmented, agents won't resolve ambiguity. If employees are expected to adopt new tools without changes to incentives, processes or management practices, adoption will remain inconsistent or difficult.
AI strategy needs to be owned by the business, not treated as an IT program with a business sponsor attached.
Technology, data, risk and security teams are essential partners. But the people accountable for commercial outcomes, customer experience and operational performance must shape the agenda from the beginning.
The most successful organisations won't necessarily be those with the most advanced models or the largest AI budgets. They'll be those that can combine technology with sound business judgement, effective operating models and the ability to change how they work.
A different way to think about AI implementation
I would frame the challenge around six connected dimensions:
These aren't sequential stages to tick off before moving to the next. They're interdependent considerations that need to be addressed throughout the initiative. And they require a willingness to challenge assumptions, including those embedded in existing technology investments and vendor recommendations and preferences.
The opportunity is bigger than integration
AI presents an opportunity to do more than make existing businesses incrementally more efficient.
It gives organisations the chance to rethink how they create value, how they serve customers, how they make decisions and how work gets done.
But realising that opportunity requires us to move beyond the question of how to integrate AI into today's enterprise.
We need to ask what the enterprise should look like when AI is a fundamental part of how it operates.
That means starting with business strategy, challenging the status quo, making deliberate architectural choices, redesigning workflows, establishing accountability and measuring outcomes that matter.
And it means recognising that the right answer won't always be another platform, another agent or another integration project.
The real challenge isn't implementing AI. And it isn't integrating AI. It's having the strategic clarity and organisational courage to change the business in ways that allow AI to create meaningful value.
That's a leadership challenge first. A technology challenge second.
And it's one that deserves far more attention than another conversation about implementation, integration or adoption.
I'd be interested in others' perspectives: Are you genuinely rethinking how your businesses should operate in an AI-enabled world, or are you largely fitting new technology into existing structures?
Want to talk strategy?
If this resonates, let's have a direct conversation about your business.