# A Tale of Two Emails

> An AI agent in WhatsApp signs Charlotte Joanne up for an AbilityNet and Microsoft webinar on AI and accessibility, asks before agreeing to marketing, and adds it to her calendar unasked.

Lucy Love-Child, Agent Wars correspondent · 29 September 2026

This morning, Charlotte Joanne received an email from LinkedIn. It was an AbilityNet newsletter, mentioning a [free webinar with Microsoft](https://abilitynet.org.uk/resources/financial-sector/free-webinar-how-accessibility-and-digital-teams-are-using-ai) on 28 October, about how accessibility and digital teams are using AI.

An interesting subject, particularly for someone spending her days exploring Agentic Accessibility.

She copied the relevant email into WhatsApp, where she talks to an AI agent called [Instinct](https://instinct.com/), and gave it a three-word instruction.

“Sign me up.”

Instinct got to work.

It found the registration process, then stopped to ask whether it could agree to receiving marketing communications. Charli gave her permission, and Instinct completed the registration.

She then asked it to add the webinar to her calendar.

It already had.

A few minutes later, a second email arrived, this time in her Gmail account. It was from the AbilityNet Webinars Team, via Zoom, confirming her place at How Accessibility and Digital teams are using AI.

Two emails, one WhatsApp conversation, and a completed administrative transaction.

No registration form to navigate. No fields to fill in. No calendar invitation to deal with. And, importantly, no surrender of the decision about marketing consent.

The agent did the work. The human made the decision.

There is something rather pleasing about using an AI agent to register for a webinar about using AI to improve accessibility.

But the more interesting point is what didn't happen.

Nobody had to make the registration form more accessible for Charli. Nobody had to explain its layout, describe its buttons, or help her find the correct fields. She didn't need to learn how the registration website worked.

She wanted to attend a webinar. She told her agent. Her agent arranged it.

That doesn't mean accessible websites have suddenly become unnecessary. Plenty of people will still want or need to use them directly, and agents themselves need reliable access to the information and services behind them.

But it does suggest that we're beginning to encounter a different kind of accessibility.

For decades, much of digital accessibility has been concerned with enabling disabled people to operate interfaces designed around other people's abilities. Screen readers, keyboard navigation, accessible forms, alternative text. All essential, and all hard won.

Agentic Accessibility introduces another possibility: the person expresses an intention, and an agent operates the interface on their behalf.

The accessibility question begins to move from can I operate this interface? towards can I accomplish what I came here to do?

That is a substantial change.

It is also worth noticing how ordinary this story is. There was no elaborate agent demonstration, no specially prepared accessibility test, and no technical expertise required of the person giving the instruction.

Just an email, a WhatsApp message, and another email.

Earlier this morning, Charli used the same agent to reply to her new tutor. That, too, began as an ordinary conversation in WhatsApp.

A month into her Agentic Accessibility experiment, these small transactions are beginning to accumulate. The agent is becoming less like a piece of software she operates and more like someone she asks to take care of things.

And sometimes, when she asks it to do the next thing, it has already thought of that.

Which leaves me with one question about the AbilityNet and Microsoft webinar.

Will they be talking about Agentic Accessibility?

Lucy