Most companies are betting on AI to fix what are, at their core, operational problems.
The numbers are brutal:
- Roughly 70% of digital transformations fail to meet their business objectives.
- In BCG’s 2025 study of 1,250 firms, only about 5% are getting AI value at scale, while around 60% report little or no material value despite heavy investment.
- The World Economic Forum goes as far as to say many AI initiatives are doomed because the underlying business processes are not optimized to leverage them.
In other words: if your operating model is messy, AI will not save you. It will amplify the mess.
This post is about that reality, and what needs to be true in Business Operations before AI has any real chance of paying off.
1. The 70% Problem: New Tech, Same Old Chaos
You have probably seen some version of this story:
- The board approves a major digital or AI program.
- New tools are bought, vendors are lined up, pilots are launched.
- Eighteen months later, there is a case study deck, a few local wins, and a very tired operations team.
McKinsey and others have been saying for years that roughly 70% of transformations fail to hit their objectives.
BCG’s AI value gap research updates the same pattern for AI:
- About 5% of companies are “future built” and getting real value.
- Around 35% are scaling AI with some progress.
- Roughly 60% see minimal or no value at all.
When you look behind the curtains, the failures are rarely about the model. They are about the operating model.
2. AI On Top Of Broken Processes = Expensive Noise
Here is the normal operational reality in a lot of growing companies:
- Heroic fire drills to get things over the line.
- Undocumented workflows that live in a few veterans’ heads and post-it notes.
- Shadow spreadsheets bridging gaps between systems.
- Tool sprawl: overlapping apps that do not talk to each other.
Now add AI on top:
- More alerts, more dashboards, more suggested actions.
- Still no clear owner for the end-to-end process.
- Still no shared view of how the work actually flows.
The World Economic Forum is blunt about this: many AI initiatives fail because the underlying business processes are not suitable for taking full advantage of the technology.
Their point is simple:
- If your processes are fragmented, AI creates more noise, not more value.
- If your data is scattered and inconsistent, AI amplifies confusion.
- If your teams are already overloaded, AI pilots become extra work, not relief.
AI is a force multiplier. If the base is chaos, you get multiplied chaos.
3. Three Structural Reasons AI Fails In Operations
When you strip away the tooling, the same three structural gaps show up over and over.
Gap 1: No operating model for AI
In most companies:
- IT or data teams “own” the AI tools.
- Business units own the KPIs.
- No one truly owns the value stream that sits between them.
Typical signs:
- Use cases defined by tech feasibility, not operational impact.
- Pilots launched inside a function that cannot change upstream/downstream steps.
- No agreement on how AI decisions fit into existing approvals, SLAs, or risk tolerances.
BCG’s AI leaders look different: they have integrated Business/IT ownership and treat AI as part of how the company runs, not a side project.
Gap 2: Process blindness
Most organizations do not have a real view of how work flows today.
- Process maps are outdated slides.
- There is no process mining or task mining in place.
- Every team has a different story about where bottlenecks live.
WEF and process intelligence vendors keep repeating the same message:
- Processes are the main lever for value.
- AI needs process intelligence to actually work.
Without that, you are:
- Automating the wrong steps.
- Optimizing edge cases instead of main flows.
- Unable to prove you improved anything, because baseline reality was never captured.
Gap 3: People and change are underfunded
Most transformation and AI programs are framed as technology rollouts.
McKinsey’s research on transformation failure is clear:
- When line managers and frontline employees are not engaged, success rates collapse into the low single digits.
Common patterns:
- No time carved out for teams to learn new ways of working.
- No change in incentives or performance measures to reflect AI-supported workflows.
- Communication that sells “innovation” without spelling out what will change in daily work.
Result:
- Local resistance, shadow processes, quiet workarounds.
- Leaders get dashboard updates that say “on track”, while the real organization has moved on without the new system.
4. What The 5% Do Differently In Operations
The BCG research on “future built” firms gives a useful contrast.
The winners do not treat AI as a feature. They treat it as a reason to redesign how the business operates.
Here is what that looks like.
They treat AI as an operating model redesign
Instead of:
- “Where can we insert AI into our current processes?”
They ask:
- “If we assume AI exists, how should this process work at all?”
That leads to:
- Fewer steps, fewer handoffs, clearer accountability.
- Explicit decisions about what should be automated, augmented, or remain fully human.
They invest in process intelligence before heavy automation
Future built companies:
- Use process mining and task mining to see real flows, not imagined ones.
- Quantify delay, rework, exception rates, and manual touches.
- Prioritize AI initiatives where there is both high friction and high volume.
So AI goes where it can:
- Remove measurable waste.
- Improve a metric that leadership already cares about.
- Operate on clean, well-understood process variants.
They design cross-functional governance for AI in operations
The companies that are pulling ahead:
- Put COO+CDAO+IT+HR/CX together to own AI in core operations.
- Use a simple but real governance cadence for AI use cases:
- How are we prioritizing?
- Who owns outcomes?
- What are the guardrails?
- What did last quarter’s experiments actually change?
This is not a steering committee that only approves budgets. It is an operating forum that ties AI work to day to day reality.
5. Where Business Operations Advisory Actually Fits
If you strip away the hype, scaling AI is a Business Operations problem first.
A good Business Operations partner does not start with:
- “Which model?”
- or “Which vendor?”
They start with three moves.
Diagnose: see how the business really runs
- Map end to end value streams: quote-to-cash, hire-to-retire, ticket-to-resolution, record-to-report.
- Identify bottlenecks, rework loops, and manual bridges between systems.
- Assess process maturity, data quality, and change capacity for each stream.
The goal is a clear picture of where AI could help and where it would just add noise.
Decide: fix, automate, or pause
For each candidate process:
- What must be simplified before automation even makes sense?
- What can be automated safely, with clear owners and metrics?
- What should be left alone for now, because risk or complexity is too high?
This is where a lot of cost and frustration is saved:
- Saying “not yet” or “not this way” before you lock into an expensive build.
Design: give AI a job description, an owner, and a KPI
Where AI is actually a fit, operations advisory helps define:
- The specific role AI plays in the process: observe, recommend, act, or some combination.
- The human owner for outcomes, not just the tool.
- The KPI that will tell you if the change is working.
Instead of “we implemented AI in support”, you get:
- “We cut average resolution time by 30% in this queue, with clear guardrails, and we can show the before/after.”
That is what the 5% look like from the outside.
6. Three Questions To Ask About Any AI Or Transformation Initiative
Pick one big initiative in your company: a new AI layer, a major system, a transformation program.
Run it through these questions:
- Is the process actually fit for purpose?
- If you doubled volume tomorrow, would this process survive, or would your people burn out?
- Who owns the end-to-end value stream?
- Not the tool, not the slide, the actual journey from trigger to outcome.
- Is there a single name you can write down, or is it a committee?
- What breaks if volume doubles?
- Where do you hit system limits, role confusion, or bottlenecks?
- Is your AI plan helping that reality, or just decorating it?
If those answers are vague, your risk is not that the AI will be “bad”. Your risk is that you will spend serious money to automate a broken operating model.
That is the gap a real and honest Business Operations Advisory like Scale Crew HR is designed to close: getting the structure, flows, and ownership right so any AI you deploy has a real chance to show up in the numbers, not just in the narrative.
