People Practice
PART III · CPulling It All Together · Tools & systems

Augment where the work already is.

The tooling question is where transformations most often buy themselves a new dependency. The pragmatic answer is unglamorous and powerful: start with what you already own, add only what genuinely earns its place, layer onto your existing systems rather than ripping them out — and weigh any new platform against the captivity it might create.

The overture named “vendor captivity” as a trap. This is where we design it out — with a tool strategy built to keep you free.

Start with what you ownLayer, don't ripOff-the-shelf, eyes openAvoid captivity
Tool-agnostic by design

The strategy outlasts any specific tool

This is deliberately not a product recommendation. The AI tooling market moves faster than any roadmap, and today's frontrunner may not be next year's. What endures is not a tool but a way of choosing tools — one that keeps capability and control inside the organisation regardless of which vendor is ahead this quarter.

So the goal isn't to pick the perfect platform. It's to adopt a sequence and a set of tests that let you move now, cheaply, on what you already have — and add more only when it clearly earns its place.

Don't ask “which AI platform should we buy?” Ask “what's the smallest tooling step that delivers value and keeps us free?”

The first question leads to a big purchase and a long lock-in. The second leads to a cheap start, fast proof, and options kept open.

The order of operations

Climb the tooling ladder — don't leap to the top

Most tooling regret comes from starting at the wrong rung — buying a big new platform before knowing what you need. Climb in order. Each rung delivers value and informs the next, and most organisations get a long way without ever reaching the top.

RUNG1
Start here · lowest cost, lowest risk

Use what you already own

The AI capability bundled into tools your organisation already licenses — the assistant embedded in your everyday productivity suite. No new procurement, no new platform, and it's already where the work happens.

Why first: a generative assistant inside the office suite people use all day — e.g. Copilot within Microsoft 365 — is often the most sensible early on-ramp for HR augmentation. Zero new tools to learn the hard way.
RUNG2
Then · configure, don't purchase

Configure and ground what you own

Point those existing tools at your own content — policies, templates, process maps — so answers are grounded in your organisation, not the open internet. Often a configuration task, not a new licence.

This is where 1·B and 2·B live: the navigator and the policy stop-line are mostly about grounding existing AI in your own instruments — capability you can switch on, not buy.
RUNG3
Only if needed · targeted addition

Add a specialist tool for a proven gap

If a specific, validated need remains that your existing stack genuinely can't meet, add one targeted tool for it — scoped to that gap, chosen against the captivity tests, never as a sweeping “AI platform.”

The discipline: you add it because a proven need demands it, not because a vendor's demo was impressive. Evidence pulls the purchase; the purchase doesn't chase a promise.
RUNG4
Rarely, and last · build or platform

Build or commit to a platform — eyes wide open

Only when scale and a clear, durable need justify it should you consider a bespoke build or a major platform commitment — and only with exit, data ownership and integration settled in advance.

The warning: this rung carries the highest captivity risk. Most organisations should reach it late, if at all — and never as step one.
The sensible early on-ramp

Why “the AI already in your office suite” is often the right first move

For most organisations, the fastest, safest first step isn't a new system at all — it's switching on the AI already embedded in the productivity tools people use every day. It's worth being clear-eyed about why it works, and what to watch.

Why it's a strong on-ramp

The advantages

  • It's where the work already happens — documents, email, meetings
  • Often already licensed, or a modest add-on — not a new platform
  • Enterprise-grade security and data handling already in place
  • Minimal change management — it meets people in familiar tools
  • Immediate wins on the everyday tasks of Parts 1·A and 2·A
What to stay alert to

The caveats

  • Convenience isn't lock-in's opposite — keep your data portable
  • Ground it in your content, or answers stay generic
  • Capability still has to be built in people — the tool alone won't
  • Review its outputs with the same discipline as any AI
  • It's an on-ramp, not the whole journey — don't mistake it for done

Microsoft 365 with Copilot is named here as the most common example, not an endorsement. The same logic applies to Google Workspace with Gemini, or whatever productivity suite your organisation already runs. The principle is what matters: start in the tools you already have.

The expensive legacy question

Layer onto what you've already paid for — don't rip it out

Most organisations run expensive, embedded core systems — the HRIS, the case manager, the records platform — that are painful and costly to replace. The instinct to “replace it all with something AI-native” is usually the most expensive and most captivity-prone path available. There is a better one.

The rip-and-replace trap

Tearing out legacy for an “AI-native” system

  • Enormous cost, multi-year disruption, huge risk
  • Trades a known system for a new, total dependency
  • Migration pain dwarfs the benefit for years
  • Locks you to one vendor's roadmap and pricing
  • The captivity trap in its purest, most expensive form
The layering approach

Adding intelligence on top of what works

  • Keep the legacy system doing what it already does well
  • Layer AI on top to read, summarise and assist around it
  • Far lower cost, far lower risk, value much sooner
  • No migration, no lock-in to a single new platform
  • Replace deliberately, in time — only if it ever truly pays

Legacy systems are rarely the real blocker. The blocker is usually that no one can get information in or out of them easily — exactly the kind of friction AI is good at smoothing from the outside, without touching the costly core.

Enterprise off-the-shelf · eyes open

How to weigh a packaged enterprise tool without buying captivity

Sometimes a mature, off-the-shelf enterprise product genuinely is the right answer — proven, supported, secure. The trick is to choose one the way you'd choose a good consultant in 3·B: for what it lets you keep, not just what it does. Run any contender through these tests.

The captivity test for any platform

Green answers keep you free. Red answers are the warning signs of a trap.

Keeps you free
  • Your data is yours, exportable, in open formats
  • It integrates with what you have, via open standards
  • Pricing is predictable and scales sensibly
  • You can leave without losing your content or history
  • It augments your processes rather than dictating them
  • Strong security and clear data-residency terms
Signs of captivity
  • Your data is locked in a proprietary format
  • Integration is closed, or expensively bespoke
  • Pricing escalates steeply as you depend on it more
  • Leaving means losing your data, history or workflow
  • It forces its process onto your organisation
  • “Switching cost” is its real business model
The discipline at the core

Choose tools the way the change chooses everything: to stay free

Every tooling decision is really the same decision in disguise — and it answers to the same test as the whole series.

The smallest step that delivers and keeps you free

At every fork, prefer the option that delivers value soonest and preserves the most freedom for later. Start with what you own, ground it in your content, add narrowly and only on proof, and treat any big platform commitment as a last resort with the exit written in advance. Cheap, reversible steps beat expensive, locked-in ones — almost always.

The same founding test, applied to tools

Does this build a capability that stays, or a dependence that lingers? A tool you can leave, whose data is yours, that augments where you already work, builds capability. A platform you can't live without, that owns your data and dictates your process, is just the rescue cycle in software form. Pick tools designed to keep you free — because a change that traded one dependency for another would have failed at the very thing it set out to do.

The tooling question, rightly answered

The best tool strategy is the one that lets you start tomorrow, cheaply, and change your mind later.

Not a grand platform bet, but a sequence of small, reversible steps — beginning in the tools you already own, and adding more only when proof, not hype, demands it. Move now; stay free.

Continue Part III
Part III · D

The Wrap-Up

The roadmap, the people, and the tools are settled. The final piece steps back: what good actually looks like, how you'll know it's working, where the journey goes next — and a close not just for Part III, but for the whole arc.

→