Stop chasing perfection too soon: why CI programmes need a “maturity pick-up point”
Many organisations start improvement programmes with the best intentions —
and then follow the wrong intuition. They
- spot an area that matters
- appoint a strong local or external leader
- bring in methods, and new tools.
The pilot becomes brilliant.
A lighthouse. A showcase.
And then… it fades.
Not because the local team failed.
But because the surrounding system drags things back:
- unclear standards
- unstable interfaces
- constant firefighting
- conflicting priorities
- and a lack of daily management discipline.
Local excellence cannot outrun organisational messiness for long.
The practical conclusion is uncomfortable, but liberating:
Before you design the “perfect” future state,
you need to understand where you are picking up the organisation.
That is why we frame Continuous Improvement programmes as a staged journey:

Not as bureaucracy.
As a sequence that prevents your best work
from being undone by the rest of the system.
The core idea:
Your organisation can only improve as fast as it can absorb change
Continuous Improvement requires three things that are often missing at the start:
- skills (people know how to improve)
- capacity (they have time and support to do it)
- stability (the process is predictable enough to steer).
If any of these are missing,
the programme doesn’t fail dramatically — it fails quietly:
- momentum drops
- benefits look “uncertain”
- and the CFO concludes CI is a cost centre.
A staged approach solves this
by matching ambition to readiness.
The three stages from ambition to outcomes
| Stage | What you’re trying to achieve | What you do in practice | Data, AI and automation (right-sized) | Change management focus |
| 1) Standardise | Create repeatability. Reduce variation. Get out of “hero mode”. | Establish basic standards, clear roles, visible work, simple routines. Roll out Lean basics and daily management. | Keep it simple: create clean data “exhaust” from the day job (timestamps, handovers, case types). Use lightweight dashboards. Use AI as a helper for drafting SOPs, checklists and training snippets. Use bots for reminders and handover nudges. Avoid heavy analytics and “smart automation” too early. | Leadership alignment on what “good” looks like. Sponsor visibility. Support middle managers who must enforce standards while teams are under pressure. Clear messaging: “standardise to free capacity, not to control people.” |
| 2) Stabilise | Make performance predictable and steerable. Build problem-solving muscle. | Roll out Rapid Problem Solving in focus areas. Run value stream improvements. Launch Lean/Six Sigma projects where needed. Implement CI governance (selection, staffing, coaching, tracking). | Now analytics pays off: Pareto, flow metrics, segmentation, control methods, process mining where logs exist. AI supports diagnosis (pattern detection, knowledge retrieval, summarising insights), not decision ownership. Automation expands: workflow rules, routing, Robotic Process Automation for repetitive tasks, stronger bot support for intake quality and compliance. | Stakeholder management becomes the main job: align incentives across silos, clarify decision rights, and protect teams from “initiative overload”. Middle managers need practical cover: capacity planning, escalation paths, and a cadence that replaces constant ad-hoc firefighting. |
| 3) Optimise | Create step-changes and differentiating capabilities (beyond “table stakes”). | Identify high-value opportunities and build new capabilities end-to-end (including support functions). Redesign operating models where needed. | This is where advanced methods belong: designed experimentation, optimisation algorithms, machine learning, reinforcement learning, digital twins, AR/VR where it truly helps. AI copilots become embedded in workflows. But always human-in-the-loop for judgement, trade-offs, ethics and customer impact. | This stage is transformation. It needs narrative, ownership and adoption at scale: capability academies, communities of practice, role clarity, governance that kills distractions fast and doubles down on what works. |
The big mistake is skipping stage one and two —
and trying to buy stage three with technology.
You can’t automate an unstable process.
But you will automate instability.
What a staged CI programme looks like in real life
A practical CI programme isn’t one monolithic rollout.
It is a set of connected “enabler tracks” that build the system in the right order:
- Lean Basics
Bottom-up standardisation and stabilisation.
Teams improve their own areas, build confidence,
and make “change for the better” visible in daily work. - Rapid Problem Solving
A focused capability to resolve recurring operational issues quickly and consistently — especially across interfaces. It builds cadence, alignment, and a repeatable improvement rhythm. - Lean Six Sigma (Green/Black Belt)
For the harder-to-crack problems where you need rigorous data-driven analysis and statistical thinking — supported by a governance system that selects, staffs, coaches and tracks projects properly. - Differentiating capabilities
Once stability exists, you can invest in step-changes: experimentation, ML, bots, advanced automation, and digital augmentation — always anchored in business outcomes, not tech excitement.
This sequence is not academic.
It is how you avoid doing “the right thing” in the wrong order.
Change management is not a side-stream —
it is the delivery mechanism
Every stage has a change challenge.
But stage one and two are where programmes usually die,
because they collide with reality:
- people are already stretched
- middle managers are caught in the middle
- priorities shift weekly
- and teams quietly revert to workarounds.
A CI programme only sticks when it is designed around adoption:
- sponsors actively remove roadblocks and align leaders
- middle managers get practical support (capacity, escalation, clarity)
- process owners are empowered to own standards and performance
- CI professionals build capability and momentum rather than “owning improvement” forever.
In other words:
The programme must build a system
where improvement becomes normal — not heroic.
Closing thought:
Build the foundation that makes excellence sustainable
If you shoot for perfection too soon,
you may get a brilliant local outcome — and lose it again.
If you standardise first, stabilise next, and only then optimise, you build something more valuable than a pilot: a company that can improve on purpose, at scale, with modern data and AI as accelerators rather than crutches.
If you want, I can also add a short “self-assessment” section at the end: 10 questions to identify whether your organisation is in Standardise, Stabilise or Optimise right now — and what the next sensible move is.
Your 30–60–90 days cookbook to make it work
First 30 days: pick the right starting point and create traction
- Don’t start with the “perfect future state”.
Start with a quick reality check: are you mainly fighting missing standards, unstable performance, or a need for step-change optimisation? - Choose 1–2 focus areas where the pain is visible
and where leadership is willing to back change in daily operations. - Make one decision explicit:
what are you trying to achieve first —
standardise, stabilise, or optimise? - Then run a short, timeboxed push to create a baseline:
a minimum set of standards, routines,
and ownership that makes the work repeatable enough to steer.
Days 31–60: build a rhythm that survives pressure
- Install a simple operating cadence: short check-ins, regular reviews, and clear escalation rules. The goal is to replace ad-hoc firefighting with predictable “beats”.
- Introduce right-sized transparency: a few trusted metrics that people actually use to run the work (not to defend themselves).
- Use digital support pragmatically: lightweight dashboards, simple workflow rules, and “nudges” that improve handovers and compliance.
- Be careful with advanced analytics and “smart automation” at this stage: if the process isn’t stable, you’ll just speed up instability.
Days 61–90: lock in adoption and expand carefully
- Validate that improvements stick under stress: busy weeks, staff rotation, handovers, exceptions. If it only works on a calm Tuesday, it’s not real yet.
- Convert early wins into reusable patterns: templates, checklists, role clarity, and a small set of rules leadership will consistently support.
- Only scale once the first area shows “sticking power”. Nothing kills credibility faster than rolling out something that isn’t stable.
- If the system is now repeatable and predictable, you can start layering deeper analytics, automation, and—only when justified—machine learning as part of optimisation.
Guardrails: what to avoid
- Don’t optimise too early. Local excellence will be dragged back by the surrounding mess if the system is still messy.
- Don’t confuse documentation with discipline. A standard that isn’t used in daily routines is just a document.
- Don’t buy “Optimise” with technology. You can’t automate an unstable process — but you will automate instability.
- Don’t overload the organisation. Fewer, closed-loop wins beat many parallel initiatives every time.
- Don’t treat change management as communication. It is sponsorship, decision rights, incentives, and capacity — made real in daily management.
- Don’t leave middle managers alone. They are the adoption layer and often sandwiched between top-down targets and on-the-ground constraints.
A simple tuning mechanism: pick four levers and iterate
To keep the programme pragmatic and adaptable.
Let each focus area pick four levers based on its real constraint:
- Process lever: make work more repeatable (clear steps, clearer handovers, fewer exceptions).
- People lever: strengthen ownership and adoption (role clarity, coaching, leadership routines — and practical support for middle managers).
- System lever: use digital support that removes friction (simple workflow rules, transparency, automation where tasks are repetitive).
- Data lever: make performance steerable with the right data — not just more data. Focus on data quality and cause-and-effect: agreed definitions, consistent capture, and metrics that reflect flow (e.g., end-to-end lead times, queue times, rework loops, ageing), not only timestamps. The goal is to move from “opinions” to “signals” that guide action.
Then run a short cycle (two to four weeks),
review what changed, and decide deliberately:
keep, adjust, or drop.
The programme doesn’t win by having the “best framework”.
It wins by building a system that learns fast —
without overwhelming the organisation.
