Don’t jump ahead: why 80% of enterprise AI initiatives stall (and how to fix the operations underneath)

11 minutes ago

5 min read

By Adam Young, Managing Partner at smart/tasking

Having spent nearly four decades building, scaling, acquiring, and transforming businesses, I’ve seen the exact same technology trap play out in every major cycle.

Organisations invest heavily in shiny new platforms, expecting tech to solve underlying performance issues. Yet time and again, they leave the underlying operating model completely untouched.

Today, the pressure on C-suite leaders to deliver an AI strategy has brought this pattern to a head. Leadership teams are setting up steering committees, signing software licenses, and running generative AI pilots. But when you look at the actual return on investment, the reality is bleak.

According to research by the RAND Corporation, over 80% of AI projects stall or fail, approximately double the failure rate of traditional IT implementations. Even more striking, MIT’s Project NANDA found that 95% of enterprise GenAI pilots fail to deliver any substantial, measurable financial return.

When these initiatives stall, the knee-jerk reaction is to blame the software. But AI isn't exposing a technology problem; it's actually exposing decades of operational debt.

For years, businesses have layered new software over broken processes, relying on manual workarounds, fragmented spreadsheets, and siloed teams. Trying to deploy AI on top of that foundation doesn't magically fix the business - what it does do is accelerate the friction.

Don’t jump ahead: you cannot automate chaos.

The pragmatic blueprint for operational readiness

Understanding where to focus first

To be clear: I am not suggesting an organisation must pause every digital initiative and spend two years fixing every broken workflow before deploying AI. That would be equally foolish.

Some broken processes absolutely must be remediated first because bad data feeds will instantly corrupt your AI outputs. But in other areas, AI tools can be used to clean up processes, structure data, and accelerate operational improvement.

The key for leadership is understanding where you are today, identifying where the critical operational drag lies, and knowing precisely where to focus your effort.

Before you scale automation across your workforce, there are four building blocks that we use to dictate whether your deployment will deliver commercial returns (or simply add operational cost):

The four prerequisites of operation readiness

1. Data structure and pipeline health

AI relies entirely on the quality of what feeds it. Most enterprises are sitting on mountains of unstructured, siloed, and sensitive operational data scattered across SharePoint drives, legacy databases, and team spreadsheets. If your underlying data layer is messy or incomplete, your AI outputs will be unreliable.

2. People, accountability, and role clarity

Automation requires explicit operational handoffs. If your teams do not have clear accountability for who owns a process, where a task originates, and where it lands, an automated tool will only confuse workflows further. Define ownership before you use automation to accelerate execution.

3. One single source of operational truth

In almost every organisation we evaluate, different departments speak different languages and measure success differently. Marketing defines a 'qualified lead' one way; Sales define it another; Finance measures it differently altogether. If your leadership team hasn't locked down unified metric definitions, AI reporting will only amplify departmental misdirection and destroy executive confidence in the numbers.

4. Pragmatic culture & leadership

While framing technology as Assistive Intelligence helps lower initial adoption panic, as leaders we must be realistic with our teams. AI will change roles, and it will replace specific manual activities. The executive responsibility is to lead that change transparently upskilling employees, shifting capacity toward higher-value commercial tasks, and ensuring technology is used to elevate human output rather than just trim headcount.

Driving commercial impact and target operating model alignmentFix what feeds the dashboard

When enterprise leaders realise their operational reporting is inaccurate or delayed, the default instinct is almost always a 'rip and replace' strategy, binning existing reporting software and spending money on yet another shiny visualisation tool or custom dashboard.

This is an expensive mistake.

In 90% of cases, the issue is broken processes and fragmented workflows feeding reporting dashboards. You don't need a new dashboard; you need to clean up the operational layer underneath your existing stack so that your reporting becomes live, accurate, and audit-ready.

(And on the rare occasion where an organisation genuinely lacks a functional visualisation framework altogether, that is where a lightweight overlay like smart/sync can step in, not to replace your tech stack, but to give you immediate reporting integrity.)

The commercial reality: demanding measurable outcomes

Ultimately, AI adoption cannot be an exercise in adopting technology for its own sake. If an initiative does not deliver clear, measurable commercial value, you are simply adding another layer of cost and operational drag to the business.

Every AI or digital transformation project must be held to strict commercial outcomes:

· Productivity & capacity: Quantifiable reduction in administrative drag and manual effort, freeing up capacity for high-value tasks.

· Margin & profitability: Lower cost-to-serve and reduced operational friction that flows directly to the bottom line.

· Revenue acceleration: Faster delivery times and elevated operational execution that drive growth and customer retention.

If a pilot can't demonstrate a direct link to one of these drivers, I'd class it as an expensive experiment.

Evaluating Your Target Operating Model (TOM)

Understanding where your business sits on this spectrum requires an honest, objective look at your Target Operating Model. You cannot optimise for AI without knowing where your process bottlenecks, data risks, and accountability gaps actually live.

This is precisely why we launched our Target Operating Model Assessment at smart/tasking.

While you can run through a self-assessment on our platform, the real value comes from looking at the reality together. We frequently run facilitated assessments and executive briefings, where we sit down with leadership teams, run through their current operating model, and show them live where the friction sits and how our tools can address it.

If you want to evaluate your operational readiness before committing further capital to automation, get in touch with me directly or drop a note to the team at hello@smarttasking.com to set up a facilitated briefing.

The execution path: moving from intent to impact

Bridging the gap between an ambitious AI strategy and real-world commercial execution requires a disciplined, empirical framework. At smart/tasking, we run this transformation through our 4I execution methodology:

· 1. Insight: We bypass surface-level assumptions to measure actual operational reality, auditing data pipelines, process health, and organisational friction across the ‘four corners of the business.’

· 2. Influence: We align executive ambition with operational constraints, establishing ISO-compliant governance guardrails and clear AI safety policies before deployment.

· 3. Intervention: We deploy focused execution pods to remediate broken workflows, clean data layers, and restructure role accountabilities.

· 4. Impact: We evidence audit-ready operational outcomes, delivering measurable speed gains, reduced cost-to-serve, and verified commercial value.

The bottom line for enterprise leaders

If your business is preparing to roll out an AI initiative, or if you are currently watching a pilot program stall, stop looking at the software. Look at the operations underneath it.

AI readiness is an empirical measurement of data health, process maturity, and leadership structure. Fix the foundational bottlenecks first, build governance into the design, and give your people the structured environment they need to succeed.

Before you invest another pound in automation, ask yourself the fundamental question: Have we fixed the operational reality, or are we just preparing to automate chaos?