A Tale of Two Seats

A Tale of Two Seats (Draft for Pineapple 27)

It began with a mistake.

A woman in Canada sent me a sales email. She had visited the Brida website, she said, and noticed the language services we offered.

Only we don’t really offer language services.

Not in the way her email imagined them.

So instead of deleting it, I became curious. The message was personal enough to look personal, but not quite personal enough to suggest that anybody had really looked at Brida.

I wrote back.

I told her that her email had caught my attention, although perhaps not in the way she intended. I suspected that at least some of the prospecting was automated, and rather than accepting her invitation to a sales call, I offered another idea.

Would she talk to me about the machinery behind the email?

How had they found Brida? How much research happened before the message went out? Where did automation help? Where did it miss? And where, somewhere between database and inbox, did an actual human being enter the process?

She replied.

And this is where it became interesting.

Yes, the original approach had been automated. Yes, the description of Brida had missed the mark. Yes, that was one of the problems with trying to scale this sort of outreach.

She was remarkably open about it.

Then she did something the system had probably not expected.

She became curious too.

She said the conversation sounded interesting. She would happily talk about what happened behind the scenes, where automation helped, where personalisation became generic noise, what they had learnt.

No demo.

No sales pitch.

Just a conversation.

For a moment, an automated cold email had achieved something nobody had designed it to do.

Two people had found something more interesting than the intended transaction.

We arranged to speak.

Then life intervened.

A week passed and another email arrived. There was an apology for the delay and a link to book a time.

For a demo.

The same demo we had already established made no sense.

The link didn’t work.

Murphy had apparently joined the sales department.

A second link arrived.

Also for a demo.

By now the whole thing had become rather wonderful.

I wrote back and reminded her that we already knew I wasn’t a customer. The interesting conversation was the one we had stumbled into afterwards.

How does this stuff really work? Where does automation help? Where does it flatten everybody into the same prospect? What happens when the human sees something the process wasn’t looking for?

I also told her I didn’t want to put her in an awkward position. If she could have that conversation professionally, wonderful. If she couldn’t, I understood.

Her answer arrived shortly afterwards.

She explained that her job was specifically to book product demonstrations. The conversation we were proposing, interesting as she found it, wasn’t really something her professional role allowed her to have.

Then came the line that has stayed with me.

“That’s just not what this seat is built for.”

There it is.

The Human Bit.

Because I don’t think she lacked curiosity.

She had already shown the opposite.

She recognised that something unexpected had happened. She understood why it might be interesting. She had even stepped outside the sales process long enough to explore it with me.

The person could see the possibility.

The seat could not.

And I’ve been thinking about that while doing something apparently rather contradictory here at Brida.

We have also been using AI.

A lot.

We are preparing the Pineapple Partnership, and over the last few days we have worked through dozens of organisations we might approach.

At first glance, that sounds suspiciously similar.

Find companies. Research them. Write emails. Contact people.

Except we seem to have spent an almost ridiculous amount of time trying to make the process less efficient.

We started with companies and discovered that companies were not enough.

People sit around tables.

So we looked for the people.

Then we realised that wasn’t enough either.

What did those people actually know?

Not their title. Not the polished sentence on the company website. Something they knew because reality had happened to them.

A ship repair company knows what happens when the drawing says one thing and the vessel standing in front of you says another.

A manufacturer knows what happens when a new machine replaces some knowledge and unexpectedly makes other knowledge more valuable.

Another company knows what responsibility feels like after the employees themselves buy the business.

Those aren’t sales personas.

They are conversations waiting to happen.

Then we discovered that France wasn’t Germany.

Obvious, perhaps. Less obvious when you are building outreach at scale.

For one group, standing, relationships, savoir-faire and the pleasure of a worthwhile intellectual exchange mattered enormously.

For another, the hour had to earn its keep. The quality had to be high. The economics had to make sense. And the fact that the conversation happened in English was not an inconvenience at all. It was part of the attraction.

So we changed the messages.

Again.

And again.

And again.

All of it AI-assisted.

Which is where the story becomes awkward.

Because the technology helping us avoid generic automated outreach may not be very different from the technology that produced the generic automated email in the first place.

The difference is what we ask it to do.

One model asks:

How little human attention can we spend on each person?

Ours has gradually become:

How much human attention can AI help us afford to give them?

That is a rather different question.

Without AI, researching dozens of organisations at this depth would be difficult. Finding the tension in each company. Thinking about who might belong at the table. What they could contribute. What they might receive. How culture changes the approach. How English fits. Which question might make somebody stop reading long enough to think:

Ach, das ist aber interessant.

The setup has taken time.

A lot of time.

But we haven’t only been writing emails.

We have been building a foundation.

Once the judgement is worked out, it becomes reusable. A new French company does not require rebuilding the philosophy from scratch. Nor does a German engineering firm. The structure remains. The human detail changes.

Setting up is slow.

Tweaking becomes fast.

That may be one of the more useful places for AI.

Not replacing judgement.

Making more judgement economically possible.

There is a distinction there we may need to get much better at recognising.

Some inefficiency is waste.

But some inefficiency is care.

Some friction is bureaucracy.

Some friction is thought.

And sometimes the quickest way to destroy value is to optimise away the moment when somebody stops and says:

Hang on.

This one is different.

The Canadian email reached me efficiently.

It misunderstood Brida efficiently.

Then, quite by accident, two humans discovered something more interesting.

For a little while, we stepped outside the process.

Then the process came looking for us.

Demo?

No.

Really. Demo?

Still no.

Eventually the woman on the other side of the Atlantic explained the problem better than I could.

It wasn’t what the seat was built for.

I may never hear from that Canadian company again.

Fair enough.

But they have already given me something useful.

A question.

Are we using AI to make people better at occupying narrow seats?

Or can we use it to make the seats wider?

Because perhaps the future of work does not depend only on what machines become capable of doing.

Perhaps it also depends on what we still allow humans to notice.