Bad news shouldn’t come from a bot: Five rules to stop automation breaking trust

The bot got there first. At Australia and New Zealand Banking Group (ANZ), more than 100 bankers learned they were out of a job not from a manager, not from HR, but from an automated email politely requesting their laptops back.

The bank apologized. The union called it “disgusting.” The headlines wrote themselves.

The damage wasn’t about the restructure itself. It was about the delivery: Out of order, out of context, and stripped of dignity. An automation jumped the queue, bulldozed trust, and turned a human moment into a systems event.

This wasn’t a quirky tech glitch; it was governance gone AWOL. When sequencing breaks down, it doesn’t matter how sleek your HR tech stack looks on a slide deck — your people still get the news from a bot.

And that’s when trust unravels, inside the company, and then online.

From years of working on digital workplace governance, I know stopping this doesn’t require a 40‑page policy nobody reads.

It requires five principles - guardrails to keep automation in its lane, and humans at the center where they belong.

Principle 1: People before platforms (sequencing beats speed)

Hard news should always follow a human-first sequence: manager → individual → written follow-up → systems.

The human conversation frames the facts, offers support, and answers immediate questions. Written confirmation then provides precision and record. Only after that should systems step in, removing access or requesting devices.

This isn’t etiquette — it’s evidence. In my work, I’ve seen research and reality align: employees trust news more when it comes from their manager or a trusted local leader, not a faceless system. That credibility is crucial when the news is painful.

The bottom line is if systems act before people do, the organization has already lost control of the story, and the relationship.

Principle 2: Guardrails, not gates (automate with hard stops)

Automation isn’t the enemy; ungoverned automation is. HR technology, IT identity systems, and security tools all need to pre-stage offboarding actions.

The risk isn’t automation itself, it’s automation without brakes. Governance means:

• Dual approval: no termination workflow moves forward without sign-off from at least two functions (e.g. HR + Legal/Employee Relations).
• “No send before” rules: emails, prompts, or access changes set up so they cannot trigger before a recorded manager conversation.
• Time-boxed windows: high-impact notices only fire during local working hours.

These are governance basics. If the system can outpace the humans, the controls aren’t strong enough.

Principle 3: Clarity over chaos (decide who decides)

Governance is clarity: Who speaks first, who signs off, and what must be logged. That clarity never sits in one silo — it has to be shared across HR, Comms, Legal, IT/Identity, Security, and (where relevant) Employee Relations.

Two habits I’ve seen work in practice:

• Tabletop runs: Walk through the process with the actual people who will deliver it. Spot collisions before they happen.
• Decision logs: Record who authorized what, when. That discipline creates accountability and a single source of truth if questions arise later.

Organizations rarely fall down because they lack tools. They fall down because they haven’t formally agreed on the order of operations.

Principle 4: Ownership is a behavior (not a metadata field)

Every communication in a downsizing (manager scripts, follow-ups, team notes, company-wide lines, external statements) needs a named human owner.

Ownership isn’t a CMS field; it’s a routine. You need up-to-date templates, local variants, translations, review dates, and escalation paths.

Without this, you get contradictory drafts, stale FAQs, and last-minute scrambles that smash into each other. Governance turns good intentions into on-time, on-message delivery.

Principle 5: Governance is a living system (not a PDF on file)

Governance isn’t a one-off document. It’s a set of routines, roles, and reviews that flex as the organization shifts. Structures change, tools get upgraded, leadership priorities move on — if the governance model doesn’t keep pace, cracks appear fast.

That means building in regular reviews: quarterly checkpoints with HR, Comms, IT, and Legal to test whether the sequence still holds. It means refreshing templates and message trees so they reflect today’s organization, not last year’s chart. And it means giving someone explicit accountability for keeping governance alive — because unattended rules age badly.

The point isn’t to generate paperwork. It’s to make sure the invisible scaffolding that keeps sensitive communication on track doesn’t quietly erode. Governance done well is boring; governance left untouched is a crisis waiting for a trigger.

Quick wins this quarter

Principles matter. But routines make them real. Here are five cross-functional moves I’ve seen make an immediate difference:

1. Run a tabletop of the separation sequence with HR, Comms, Legal, IT/Identity, and Security. Walk the end-to-end process, spot gaps, and agree the order of operations before you need it.
2. Insert a sequencing gate to your HRIS/IAM flow: no badge action, device prompt, or email goes until a manager conversation is logged. Automation follows people, not the other way around.
3. Refresh your manager pack with scripts, FAQs, EAP/outplacement links, “what not to say” guardrails. Review and update it quarterly so it’s relevant to today’s org, not last year’s.
4. Build a blackout calendar for sensitive events: earnings windows, legal filings, regional holidays. It’s a governance tool that keeps messages from colliding with external obligations or local realities.
5. Schedule a quarterly review. Bring HR, Comms, IT, and Legal back together to revisit decisions, update templates, and check whether processes still fit current structures and tools. Governance isn’t static — it’s a living system.

Common traps to avoid

Even with the best intentions, organizations stumble in the same places. Governance falls apart not because people don’t care, but because habits and pressures overwhelm process. Watch for these traps:

• Speed theater: Confusing “fast” with “good.” Rushing the sequence might look decisive, but all it really does is leave a trail for you to clean up.
• Shadow systems: Side-channels, spreadsheets, or unofficial tools that bypass agreed sequencing. If it can’t be governed, it shouldn’t run.
• Executive override: Last-minute requests that scramble the order, add “one more message,” and confuse audiences. Guard the sequence with data and agreed rules, not just hierarchy.
• Stale playbooks: Applying last year’s plan to this year’s reorg, even though structures, tools, and leaders have changed. Update the RACI, the message tree, and the timing windows. Governance must evolve with the organization.

Make it boring, on purpose

When communication governance works, nobody notices. That’s the point. No headlines, no frantic screenshots, no managers ambushed by their own team’s questions. Just the hardest news delivered by humans, in the right order, with systems quietly tidying up afterwards.

The ANZ fiasco is a reminder: automation will always move faster than empathy — unless you slow it down on purpose. That doesn’t take a monster playbook. It takes three things: sequence first, guardrails always, people before platforms.

Because when governance fails, the message employees hear isn’t just “your job is gone”. It’s we didn’t care enough to get this right. And once that thought takes hold, no apology, counselling line, or press release will undo the damage.