Talks like me: how Microsoft Cowork learns your team's tone of voice

For businesses rolling out Microsoft Cowork, Talks like me is often the moment people stop seeing AI as a novelty and start trusting it with real work. Here is how it works, and why tone matters more than most people think.

Why generic AI drafts get ignored

Most AI assistants produce competent text that no one wants to send. The tone is too formal, the structure is too neat, and the vocabulary belongs to no one in particular. Recipients can spot it instantly, and senders end up rewriting almost every line. The promised time saving disappears.

Tone is not a cosmetic detail. It is how your customers recognise you, how your team builds trust, and how your brand stays consistent across hundreds of small interactions every week. If an AI cannot match it, it cannot do the job.

What Talks like me actually learns

Talks like me is a Cowork skill that builds a private style profile from your own writing. With the right permissions in place, it looks at how you phrase things in past emails, documents and chats, then captures patterns such as sentence length, formality, common openings and sign-offs, vocabulary preferences and the level of detail you typically include.

That profile stays inside your Microsoft 365 tenant. It is not shared with other organisations, it is not used to train public models, and it respects the sensitivity labels and access controls you already have in place.

What changes day to day

Once Talks like me is active, the difference shows up in small but compounding ways:

Email drafts arrive already sounding like you, so a quick read and send replaces a full rewrite. Meeting follow-ups match the way you usually summarise actions for that client. Internal updates land in your normal voice instead of corporate filler. Long documents keep a consistent tone even when several people contribute.

The cumulative effect is what makes Cowork feel different from a standard AI chat experience. It is not just generating text, it is generating your text.

Tone for teams, not just individuals

Talks like me works at the team level too. A support team can keep a warm, problem-solving voice across every reply. A finance team can stay precise and measured. A sales team can sound confident without sounding pushy. New starters benefit immediately, because the house style is baked into the drafts they receive instead of something they have to learn by osmosis.

Where it fits in a wider Cowork rollout

Talks like me is one of several Cowork skills, but it is usually the one we recommend turning on first. It delivers a visible win quickly, builds confidence in the rest of the platform, and pairs naturally with skills that handle inbox triage, meeting prep and document drafting.

As part of a broader AI programme, it is also a useful proof point for stakeholders who are nervous about AI sounding off-brand. Once they see drafts that read like the team wrote them, the conversation moves on quickly.

Getting it right from day one

The quality of Talks like me depends on three things: the right data being in scope, the right governance around it, and the right rollout plan. That means clear sensitivity labels, well-scoped permissions and a phased approach that starts with a small group before expanding.

Done well, the result is AI that finally sounds human - because it sounds like your humans.

Want to see Talks like me in action for your team?

Speak to Axon about a Microsoft Cowork pilot.