The Gauntlet
Be the firm everyone can say yes to.
You get one shot at a pitch. Make it your best shot.
The most expensive way to find out what a client will agree to is to bet everything on the best guess your team can make, and test it on the client. If the guess was wrong, you find out when the client says no, or when another firm wins the work with a better guess.
We think no line costs a firm more than the one it didn’t know would stop the client. Right now, sending your case and testing it are the same thing. Send once, test once.
The Gauntlet, from Taleist, separates the two. It tests every version of your document on Client Proxies: AI models we configure to reason as someone in each role that will decide, such as the head of the team, procurement, finance or whoever signs it off. Test it as many times as it takes, then send it once.
How to test a pitch before it goes to the client
Each role gets several Client Proxies. Every version still in play is read by all of them.
To test a pitch before it goes to the client, put it in front of the roles that will decide, and find the reason each one would say no. A colleague playing the client can’t do that. A colleague knows your firm, so they can’t read the pitch cold, and they aren’t the client’s people who sign it off. The Gauntlet does it with Client Proxies for each role.
The Gauntlet builds Client Proxies for every role that decides, including the roles that sign off without ever taking the meeting. Each Client Proxy reasons as someone in one role: not your client’s CFO, but the CFO of a mid-sized bank under pressure to cut costs.
Each Client Proxy starts from what’s known about the role: the conditions in the sector, the pressures on people in that job, and your team’s general experience of people who hold it. It’s never built from your client files, from anything a client or prospect has told you, or from anything about a named person or organisation.
Then we vary the role across the things nobody can know in advance. A procurement lead under orders to cut suppliers, or one who isn’t. A head of team who decides alone, or one who takes the decision to their own boss. Each Client Proxy is set a different way, so each reasons its own way.
Your document goes to each role’s Client Proxies in turn, in the order those roles would meet it. Where a Client Proxy says no, we try different replies on your behalf to see what turns that no around. A version with a reply that works moves on to the next role with the reply built in. A version one role stops goes no further.
The goal is to find the version no Client Proxy stopped.
The same pitch, three ways
Say you’re pitching a second service to a client. The Client Proxies for a head of operations like the idea: one firm for both, fewer pieces to manage. The Client Proxies for a procurement lead stop it: too many eggs in one basket.
A second version changes the offer. The new service will be run by a separate team, and the client can cancel it without affecting the first service. Procurement’s Client Proxies are satisfied: being able to cancel makes the risk acceptable. But now the head of operations’ Client Proxies say no: two teams is no longer fewer pieces to manage.
A third version keeps the right to cancel, but goes back to one team. The head of operations has fewer pieces to manage, and procurement can still cancel. Neither role’s Client Proxies have a reason to say no.
Three versions of one pitch, and which role’s Client Proxies stop each.
From inside your firm, you’d probably write one of those three versions, and judge it from your side, not theirs. The Gauntlet tests all three and shows you the one no role stops.
We ran it on our own launch first
Before we launched a new product live to a room with more than 200 possible buyers, we ran The Gauntlet on its sales page. Three roles decide whether a firm buys it: the director of business development, the chief operating officer and the managing partner.
We’ve written direct-response copy for 30 years. What those years teach you is that your instinct doesn’t matter. What the buyer thinks does. Until now, there was no way to put a draft in front of the buyer. You found out whether you’d got it right on launch day.
Seven choices on our page change how those three roles read it, from the headline to a single line about IT. With two or three options for each, and nine possible opening lines, that makes 1,728 different pages. The draft we wrote was just one of them.
Here’s what it took to test them before launch:
The result was one page. It wasn’t our draft.
Experience writes one of the 1,728, sending the client a best guess. The Gauntlet searches all 1,728 to find the positioning no Client Proxy stops.
Why not ask ChatGPT?
One prompt to ChatGPT gets you one opinion of one version. Ask it again and you get variations on the same answer, because a different answer needs a different set-up: a different Client Proxy, set differently on one dimension. You also need to keep each Client Proxy separate, so one answer doesn’t contaminate the next, and a record of which setting produced each response. On our own page, that came to 471 readings, run and recorded.
What happens, and what you get
- Fill in a short questionnaire
Describe the sector broadly enough to fit several organisations. Tell us the roles that decide and the order they’ll see your document, and your team’s general experience of people in those roles. Send the document and the versions you’re considering, with anything that describes a real client rewritten to describe the sector instead. No client files, emails, call notes, CRM records, invoices or fee disputes, and nothing a client or prospect has told you. We don’t need to know who the client is, and we don’t record it.
- Take an intake call
We’ll ask what’s decided, what can still change, which roles decide, which sign off and which would argue for doing nothing. We talk about roles, not people, and we won’t ask who the client is.
- Come to the debrief
What we do
- We work out which roles decide: which are in the meeting, which the document gets forwarded to, and which sign off without ever taking the meeting.
- We build Client Proxies for each role from the sector, the job and your team’s experience, and vary each role across the dimensions that matter for it.
- We test every version of your document, different ways of opening it, and a set of replies to each reason to say no, role by role, in the order they’ll meet it.
- We test what happens when the document is forwarded upstairs: what the next role still sees, what drops out and what turns into a promise you never made.
What you get: a report, as a web page and a PDF
- The work, counted, on the first screen. How many readings, how many words of feedback, and the version no Client Proxy stopped.
- The version to send, word for word, and every opening tried.
- Your document, before and after. What it said, what it says now and which role’s Client Proxies made the difference.
- How it travels. What each role takes from it, what it passes on and what it still needs.
- If they say this, say this. Every reason to say no the Client Proxies raised, the reply that persuaded them, and how each reply went.
- The price. How to put your fee, if your document carries one.
- In the meeting, find out first. The questions that answer what nobody could know in advance.
- Dated predictions about what will come up in the meeting, which we score with you afterwards.
- A named person at Taleist to call before the meeting.
When to use it
For moments with a date on them and money on the line: a pitch, a change to fees or terms, a key-account plan, a new service or the plan for a strategy day or a partners’ meeting. For a plan, the Client Proxies are the clients it’s aimed at, not your partners.
See it for yourself
The demo shows you the report from our own Advising Upwards page, with the animation: every version, every reason to say no, every reply tried and the dated predictions we’ll score after the page goes live. We’ll also show you a sample report on a fixed-fee proposal, made up from start to finish to show the format.
Nothing in the demo involves your work. If you like what you see, the next step is a test you can judge yourself. We run The Gauntlet on a pitch whose outcome you already know, won or lost, with anything that describes the client rewritten to describe the sector. You compare what the Client Proxies raised with what you remember the real client raising. You don’t need to tell us what the client said.
Questions
Will it tell us whether the client will say yes?
No, and nothing honest can before the meeting. The Gauntlet tells you which role could stop your document, why and which replies persuaded each role’s Client Proxies. That’s something you can show your partners, with no probability in it to argue about.
Does it test our price?
It tests how you put the fee to the client, never the number. A Client Proxy asked about a figure says no because it was asked, so a tested figure would tell you nothing. The figure is your call.
Is this a digital twin?
No. A digital twin is built to play a named person. We don’t build those, and we’d suggest you don’t either. A model built to play a named person, from what you know about them, is considered personal information by regulators, which means you need the person’s permission to model them. For law and accounting firms, much of what you know about a client’s people is also your client’s confidential information. We aren’t lawyers, and this isn’t legal advice, so you should speak to your lawyer about your situation. However, we do not build digital twins. We build Client Proxies for roles, not people, and vary each role across the things nobody can know in advance.
How is it different from ASI?
ASI is Taleist’s market research service. It configures hundreds of Client Proxies, across every dimension that matters in a market, and asks them many questions. It tells you how a whole market thinks, and how widespread each view is. The Gauntlet configures a few for each role in one decision, on only the dimensions that matter for that role, and asks one question: how do they reason about your document? Its findings are reasons, replies and wording, never counts of who agreed. Put new terms to all your banking clients, and that’s ASI. Put new terms to one bank, and that’s The Gauntlet.
How do you know it works?
Every Gauntlet report sets dated predictions, and we score them with you after the meeting. That’s how The Gauntlet builds its record. The first report, on our own launch, is scored after the launch. If you want evidence on your own ground first, we run The Gauntlet on a pitch whose outcome you already know, and you judge whether the Client Proxies raised what the client raised.
What do you need from us?
For the demo, nothing. To run The Gauntlet on your own pitch, offer or message: the sector, described broadly; the roles that decide; your team’s general experience of people in those roles; and the document with the versions you’re considering.
Do we have to send you confidential material?
No, and we won’t take it. No client files, emails, call notes, CRM records, invoices or fee disputes, and nothing a client or prospect has told you. You don’t need to tell us who the client is, and we don’t record it. Everything we take in and send back is text. If your document describes a real client’s situation, you rewrite those details to describe the sector before you send it.
What happens to our information?
It reaches us already anonymised: before you send anything, you take out the client’s name, the names of its people and anything else that would identify either. We keep what you do send us in Taleist’s own systems, siloed from every other client’s work, and use it for your run and nothing else. The Client Proxies run on Anthropic’s Claude and OpenAI’s models, through their business services, and neither trains its models on what we send. We delete your material 90 days after we hand over the report.
Book the demo
See The Gauntlet run on our own launch
Book a demo and we’ll walk you through the report from our own launch and the sample report, version by version. Nothing in it involves your work.
Prefer email? asi@taleist.agency
Tell us the pitch or decision in front of you and we’ll reply within one business day.