Tell Your Team Before You Tell the Market: Internal Communications and AI for Small Business

A small business owner with a press release ready to go, proud of the announcement and planning to send it to journalists that afternoon, should ask one question first: does the team already know?

If the answer is no, or worse, “they’ll find out when everyone else does,” the order is wrong. A new partnership, a product pivot, a change of premises, an ownership change — these are the announcements small businesses spend real money and time getting right for customers and media. Then staff hear about it from a client, a LinkedIn post, or a journalist ringing to ask how they feel about it.

That’s a sequencing mistake more than a communications failure in the usual sense. AI tools are genuinely useful for fixing it. Most small businesses already have the goodwill toward their staff; what’s missing is the time to draft a second version of every announcement before the first one goes external. This post covers the workflow, the real costs, where AI runs out of road, and what to do instead when it does.

Why the Order Gets Reversed

Press releases and social posts have deadlines attached: an embargo, a journalist’s interview slot, a launch date tied to a trade show. Internal announcements rarely have a deadline that feels as sharp, so they slip to “I’ll tell the team this afternoon” and then afternoon becomes the next morning.

By the time that happens, the external version is often already circulating. A supplier mentions it in a group chat. Someone shares the LinkedIn post. A team member who should have heard it from their employer first instead hears it from a stranger, and the content of the news stops being the story. How they found out becomes the story.

Finding out secondhand can do more damage to staff trust than the news itself. A hard message delivered directly at least shows respect. Hearing it from a stranger suggests the business didn’t think its own people were worth telling first.

What Counts as a “Major Change” Your Team Should Hear First?

A major change is any announcement that affects how someone’s job works, where they work, who they work for, or whether their job still exists. That includes redundancies, restructures, relocations, mergers or acquisitions, leadership changes, and significant shifts in products or services that change day-to-day roles.

If customers, media, or investors are learning something that changes an employee’s working life, that employee has earned the right to hear it first, in plain language, with a chance to ask questions. A general product update or minor hire doesn’t need this treatment. A pivot that makes someone’s current role redundant absolutely does.

A simple test: if a staff member would reasonably feel blindsided reading this on LinkedIn, it needs an internal version first.

The Workflow: Drafting Internal Announcements With AI in Under an Hour

This workflow suits a business that already has an external announcement drafted and needs the internal version fast. It assumes you’re using a general AI chatbot such as Claude or ChatGPT, not a dedicated HR platform.

1. Brief the AI with the facts, not the spin (10 minutes). Paste your external press release or announcement draft in as reference material, then separately list what staff specifically need to know: what’s changing, when, who it affects, what stays the same, and who to ask questions. AI drafts worse internal messages when it’s only given the polished external version to work from.

2. Generate a first draft (5 minutes). Ask for a direct, plain-language internal message, not a scaled-down press release. Internal announcements should sound like a person talking to colleagues, not a company talking to the market.

3. Draft for the specific audience, not everyone at once (10–15 minutes). A restructure affects different teams differently. Ask the AI to produce a short variant for the team directly impacted and a separate, shorter version for everyone else. One generic email to the whole company for a change that only affects one department reads as either careless or evasive.

4. Strip the AI tells and add real detail (10 minutes). Replace vague AI-generated phrases like “we’re excited to announce” or “this reflects our ongoing commitment” with actual specifics: dates, names, next steps, where to go with questions. Vague language is the fastest way to make staff suspect you’re hiding something. Whether to tell staff that AI helped draft the message is a separate question, covered in AI Wrote the First Draft. Do You Have to Say So?.

5. Get a second human read before you send (10 minutes minimum). A co-founder, HR adviser, or trusted manager should read it for tone and for anything legally sensitive before it goes out. This step isn’t optional for anything touching employment status.

A copyable starting prompt:

“Here’s our external announcement about [change]. Write a direct, plain-language internal message for staff explaining what’s changing, when, how it affects their day-to-day work, and who to contact with questions. Avoid corporate phrases like ‘exciting news’ or ‘we’re thrilled.’ Keep it under 200 words. Then write a separate, shorter version for staff not directly affected by this change.”

Total time for a routine announcement: around 45 minutes, most of it in steps 3 and 5. For anything involving job losses, double that, because the review step needs to be slower, not faster. If the change is a crisis rather than a planned announcement, the external playbook is in How to Use AI to Write a Crisis Communications Plan (Before the Crisis Hits).

Can AI Write a Redundancy or Restructure Announcement?

AI can draft the structure and tone of a redundancy or restructure message, but it can’t tell you whether that message meets your obligations under the relevant award, enterprise agreement, or the Fair Work Act, and it shouldn’t be the final check on anything touching someone’s employment.

Many awards and agreements include consultation requirements for major workplace change, and the specifics vary by industry and by the agreement your business operates under. Check the Fair Work Ombudsman’s guidance or talk to an HR or industrial relations adviser before the message goes out, not after. AI is a drafting tool here, not a compliance check.

What AI is genuinely useful for in this situation: getting a first draft of the general announcement to the wider team out of the way quickly, so a trained adviser’s time goes into the harder, individual conversations rather than wordsmithing a group email.

What Not to Do When AI Drafts the Message

These mistakes usually come from rushing rather than bad intent.

Don’t use the same AI prompt and tone for a redundancy announcement that you’d use for a product launch. The cheerful, benefits-forward language that works for marketing reads as tone-deaf, sometimes cruel, when the news is bad.

Don’t send the AI draft without a human edit pass. Generic phrasing in an external post is forgettable. In a message about someone’s job, it reads as a company that couldn’t be bothered to write to its own people properly.

Don’t let AI draft content for individual, difficult conversations. A redundancy conversation with a specific employee needs a manager in the room, not a script. Use AI for the general announcement and the FAQ document, never for the one-on-one.

Don’t skip the timing check. If your external announcement is scheduled for 9am, your internal version needs to land well before that, with enough lead time for questions, not five minutes before the press release goes live.

Tool Comparison: Chatbot vs HR Platform vs Template Library

For most small businesses under 20 staff, a general AI chatbot (Claude or ChatGPT) with a clear brief is the fastest and most flexible option, and it’s what the workflow above assumes. The output needs more editing than a purpose-built tool, but it adapts to any situation rather than forcing your announcement into someone else’s template. How to Brief AI Like a PR Professional covers what a clear brief looks like.

HR platforms with built-in communication features, such as Employment Hero, include announcement templates alongside the compliance paperwork for things like restructures. These are worth the subscription if you’re already paying for the platform for payroll or HR records, since the templates are built with Australian employment law in mind. They’re less useful if you don’t already have the platform, because you’d be paying for a full HR system to get a feature a chatbot can approximate for free.

A static template library (a folder of past announcement drafts) is the cheapest option and works fine for routine, low-stakes updates. It falls over for anything unusual, because templates don’t adapt to specifics the way a properly briefed AI draft does.

What Does Getting This Right Actually Cost?

A general AI chatbot subscription such as Claude Pro runs around US$20 a month billed monthly, correct at the time of writing on claude.com’s pricing page — check current pricing before you commit, as plans and prices shift. If you’re already using one for other PR or marketing work, there’s no additional cost to use it for internal comms.

Budget separately for the human review step. An HR or IR adviser’s time for a sensitive announcement, including sign-off on a redundancy or restructure message, varies by adviser and by how complex the change is, so get a quote before you commit if you don’t have in-house HR. That’s not optional spend for anything touching employment status; treat it as part of the cost of the change itself, not an add-on.

The cost of getting it wrong is harder to put a number on but larger in practice: a Fair Work claim, a team that stops trusting internal announcements, or a resignation from someone who finds out secondhand. None of those show up on an invoice, which is exactly why they get underweighted when businesses are deciding where to spend an hour of drafting time.

What AI Can’t Do Here

AI can’t read the room. It doesn’t know that your ops manager is going through a divorce this month and needs a softer delivery, or that last year’s restructure went badly and this one needs extra reassurance up front. That judgement sits with the person delivering the message.

AI can’t make a legally compliant call on consultation timing, entitlements, or process. That’s a job for an adviser who knows your specific award or agreement, not a chatbot working from general knowledge.

AI can’t have the actual conversation. A draft email gets the words on the page. The difficult, individual conversations that usually sit alongside a major announcement still need a person, in the room or on a call, able to respond to what’s actually said back.

And AI can’t take responsibility if the message lands badly. That stays with whoever signs off on it, which is exactly why the human review step in the workflow above isn’t optional.