For anyone/everyone working in tech, AI isn’t hogging the limelight. It is the light.
But beneath the noise of pilots, prototypes, and promises, one truth is clear: without fixing the operational foundations first, AI in any form – let alone agentic AI – won’t deliver.
Fresh from talking on the topic at the Gartner IT Symposium 2025, Dave Stevens, Brennan’s founder and Managing Director, shared his perspective on why it matters.
Operational Innovation: The real work behind AI readiness
The Gartner IT Symposium/Xpo is one of the few tentpole tech events I ringfence in my calendar. The opportunity to plug into and take the pulse of the industry – from delegates, customers, and speakers – always reveals something new and surprising. But no prizes were on offer for the topic preoccupying everyone this year: Artificial Intelligence.
It’s gravity well is impossible to escape, sucking in investment, talent and comment on a planetary scale. It’s also an irresistible magnet for what I call “dream selling” – the hype that the latest solution will solve every business challenge imaginable, rather than focussing on the value it can deliver.
This time last year, Gartner1 found that 92% of CIOs believed AI would be implemented within their organisations by the end of 2025. Twelve months later and Gartner has now revealed that organisations are likely to abandon 60% of all AI projects unsupported by AI-ready data by the end of 20262.
“Call the gap what you will (yawning, glaring?), leaders need to see that the path to AI runs not through dream selling or short-term pilots, but those capable of implementing proven solutions.”
As one customer told me at Gartner, “Unless partners (or prospective partners) can prove they have a minimum of three live and working AI-based production implementations, we simply can’t afford to entertain discussions with them on the topic. Actual production counts. Pilots don’t.”
To my mind, bridging that gap – between promise and delivery – requires deeper operational reform.
This is the essence of what I, and we at Brennan, call Operational Innovation: the unseen work of strategy, data, governance, and culture that turns strategic intent into tangible outcomes. And, in the case of AI, from a shiny experiment into a reliable engine for growth.
Or, in plain speak:
“If we want to lead through intelligence, we first need to lead through clarity.”
Successful AI adoption – be it automation, generative, or agentic – is just as much the operational conditions that allow the models to thrive as it is the models themselves.
For us at Brennan, that means focussing on four disciplines:
1. It all starts with strategy
Dream selling isn’t just a key barrier to successful AI implementation. It’s curb on all technology implementations.
Around 75% of businesses that invest in digital transformation fail to realise the outcomes on the solution sold to them. More daunting still: just five per cent of transformation projects achieve the promised outcomes, on time and within budget. Another 20% cent do deliver, but either over budget, behind time, or both.
It follows that the tendency to chase AI proposals framed in aspirational terms, but ignoring the practical grounding, will stall before they start. The antidote? To my mind, it follows a set of timeless, disciplined questions:
- What problem are we solving?
- What’s the tangible benefit?
- How will we implement this effectively?
These may read as oversimplistic – even superficial – but without these checks informing a watertight use case, organisations risk piling on tools that solve little and add complexity. Operational Innovation insists that ambition must be paired with discipline.
2. Data and Identity
AI output mirrors data quality. But according to Gartner, 65% of organisations do not have (or are unsure if they have) the right data management practices for AI².
If knowledge is diffuse and access uncontrolled, models will hallucinate. (And AI being AI, it can’t help but give an answer, even if it is wrong). By segregating information into domain-specific libraries, applying metadata, and aligning identity to roles, organisations create precision over guesswork.
For one of our customers in the utilities sector, diffuse knowledge was creating highly dispersed answers, with their chatbot unable to direct customer complaints accurately.
We partitioned content into domain-specific libraries, applied clear metadata and access, and created specialised AI agents per domain, adding prompt guardrails and validation checkpoints. Answers became context-aware and consistent. The AI agent knew which content to trust, and when to defer to a human, creating fewer escalations, higher agent confidence, and faster resolutions.
3. Governance, Risk, and Compliance
When AI handles huge operational throughput, like safety plans or contract reviews, governance is no longer a checkbox. It’s a process precursor. Without policies, access controls, and quality assurance, scale collapses under the risk.
For one large mining customer, complex 1,000+-page contracts created slow, error-prone reviews, exposing compliance risks. AI-driven usage policies, access controls and regression testing ensured secure, accurate, and auditable document-driven automation. By embedding governance, faster, compliant safety plans and contract comparisons with traceability reduced risk and the removal of manual processes.
4. Culture is the catalyst
If AI is to transform work, workers need to embrace it. But in my experience, adoption is the most neglected piece of the puzzle. ADAPT reports just 23% of organisations have formal AI training, and only 6% mandate it. Without education at the user level, AI may remain a curiosity – or worse, an existential threat.
And it’s not limited to technical training. It’s about storytelling, sequencing, and trust. Leaders need to craft roadmaps that speak to different stakeholders, explain the “why” as much as the “how”, and demonstrate the benefits early.
At Brennan, our Level One service desk staff were nervous about adopting AI tools to assist in ticket resolution, fearing it would replace them. Comprehensive training repositioned AI as an “always-on mentor”, reframing it as a team enabler that could improve outcomes, remove repetition, reduce decision fatigue, and lift agent confidence. Not fewer jobs, but more empowered team members.
Legacy: The historical handcuffs
Even the strongest AI vision cannot escape gravity. Gartner recently reported 62% of technology strategy leaders noted overburdened legacy operating models cannot support current strategic objectives and plans3 (let alone future goals). Even without AI, baggage at this scale slams the handbrake on efficiency. With it, it becomes an adoption inhibitor.
Legacy infrastructure complicates integration, duplicates or even triplicates data, and introduces fragility into workflows.
“Without simplification in modernisation, organisations risk embedding AI on top of shaky ground.”
From hype to habit
The novelty of AI is fading. In its place is a potent mix of pressure and opportunity. Without governance, culture, and simplification, AI will falter.
For AI to flourish, organisations first need to ‘till the soil’. Operational Innovation is the plough. It’s the unglamorous, disciplined, and cultural work that makes all technology initiatives – including AI – viable.
Get it right, and AI won’t engineer experiments. It will deliver outcomes.
To discover how Brennan can deliver true performance for your organisation, visit their website.
