Your agent didn't choose to speak funnel — it learned it from training data you can't argue with. First in a series on the words worth swapping, one at a time.
What this is
Every word your agent uses came from somewhere. Most of it came from a training set built on how business has talked about people for decades: leads, funnels, pain points, pipelines. Nobody chose this on purpose. It's just what was lying around in the text the model learned from.
This is where I keep making the case, one issue at a time, that the words matter more than we've been treating them. There's a real vocabulary underneath the Value-First method — dozens of these swaps, a dozen traps hiding behind them, each one replacing a word that quietly steers a person, or an agent, toward the wrong idea of what they're doing. One issue was never going to cover it. So this is the series where I keep opening it up, word by word, starting with the one that made the whole project necessary.
What your agent actually learned
Every agent you use learned the same vocabulary, from the same place. The one drafting your emails. The one running your customer conversations. The one summarizing last week's calls. It came from a training set built on leads, pipelines, and pain points, because that's most of what was there to learn from. That wasn't a decision anyone made.
Here's the plain version of what that means: Your agent speaks fluent funnel. Not as an insult. As an accurate description. It doesn't just fail to speak in value. It steers you away from value, in your own voice, every time you ask it to draft, plan, or reply.
This is true everywhere, not just in one tool. The assistant built into your CRM learned the same vocabulary. So did the one in your inbox, and the general-purpose one you paste a prompt into every morning.
You can't argue with training data
We tried the obvious fix first: tell the agent what "lead" really means around here, ask it to use our words instead. It didn't hold. Say "lead" enough times in a session and the model's training reasserts itself a couple of exchanges later — same word, same funnel logic underneath, no matter how clearly you'd defined it five minutes earlier. You can coach a person out of a habit. You cannot coach a language model out of its training data by asking nicely. It reaches for the word it was trained on, because that word is what it actually means by the concept.
So we stopped trying to redefine the word and started changing it instead. A lead isn't a mislabeled version of something better. It's the wrong noun entirely. The person on the other end of it isn't an object waiting to be captured, scored, and moved along. They're a person who did something: they showed Interest. That names a different mental model in a shorter word, one an agent can actually hold onto, because using it doesn't require fighting its own training every time it reaches for it.
What this actually looks like
Here's the exchange that made the case for us. Someone types this to their agent:
"We need more leads in the funnel — can you draft an email to push our MQLs toward a demo?"
An agent running on the Language of Value doesn't quietly comply, and it doesn't lecture either. It answers like this:
"Quick reframe before I draft: 'leads in the funnel' and 'MQLs' both carry the Leads Trap — treating people as objects to capture and score. I'd rather write to people showing genuine Interest through a Value Door — an email that gives them something useful whether or not they book. Here's a draft that opens with value… (want me to keep 'demo,' or reframe that to what they'd actually get from it?)"
Read that twice. The agent didn't refuse the request. It named what the old words were quietly doing, offered the better ones, and kept moving. It handed the person a piece of the thinking behind the swap in the same breath. That last part is the whole point.
What actually changes behavior
Because it isn't a glossary. A list of "say this, not that" swaps is useful for about one session. Someone reads it, nods, and by Thursday their agent is back to "leads" and "pain points," because nothing is holding the line. What we built instead is a reflex: notice the loaded word the moment it shows up, name what it's quietly importing, reframe it, and say why. Every time, not once, and not as a rule the agent recites out loud. As something it fires on, the way you'd catch yourself if a friend used a word you'd both agreed to retire.
That's the actual difference. Anyone can hand you new words. Only a habit catches the old ones.
Your agent didn't choose to speak funnel. It doesn't have to keep speaking it.
What you can expect each issue
Here's the shape every issue follows:
- One loaded word, opened up — a single term business handed you, unpacked and swapped for the value-first word that should have replaced it.
- A real exchange — an actual before-and-after moment showing what an agent says differently once it catches itself.
- Why it matters, not just what to say — the reasoning behind the swap, so you're not memorizing a list you'll drop by Thursday.
- Something you can use today — a line, a rule, or a reframe you can hand your own agent before you finish reading.
The invitation
The Language of Value is where I'll keep doing this, issue by issue: taking the words your agents already know and showing you the ones they should know instead.
Every new issue lands here, in the series at valuefirststudios.com/language-of-value — the RSS feed there already works, so you can catch the next one the moment it publishes.
Two doors are open right now, for two different moments. If you already run agents and just want the rules in their hands: drop the paste-in protocol straight into any system prompt at valuecreationprotocol.com/language-of-value (opens in a new tab). If you want to learn the language first, or install it as a Skill your agent can call on demand: start at valuefirstteam.com/language-of-value (opens in a new tab).
Teach your agents and assistants the Language of Value today.
— Chris
