Over the past three weeks my team worked through a frantic pitch cycle. After a decade in this business, writing proposals is still mentally and physically punishing — and still where the most learning happens. Stepping out of routine operations to think about new activity and plans grows you fast. It is probably why senior people in the agency world say the pitch is the flower of the business.
There is no correct answer in a proposal. An RFP shares the requirements, but interpretation is wide open, and every agency builds its own answer. You usually get two to three weeks to cover RFP interpretation, writing, design, and video work at once — a tight schedule made tighter because most staff carry regular client work alongside it. Most of that period gets spent working late.
This pitch was no exception. But the density of the work was different, because of AI.
Ideas are the core of a pitch. Eighty to ninety percent of the work is validating in meetings how well the core concept fits the RFP and sharpening the idea further.
But not everyone brings the same density of thinking and information. Naturally, most concepts end up heavily reflecting one or two key people's opinions. If the key person's idea looks plausible, the whole team can slide into confirmation bias around it.
AI reduces that tendency. It evaluates ideas without deference or preconception and checks fit against the RFP. In this pitch, I threw a promising meeting idea at it and got back: this idea does not match condition X in the RFP. Coming from a person, that critique might have landed as opposition or been taken personally; coming from AI, it went on the table with relatively little emotion. That single note redirected the proposal at one point.
Pitch meetings are full of things almost-but-not-quite grasped. When discussing campaign structure or format, if nobody diagrams what is in their head, everyone understands something different. Logic that sounds coherent spoken aloud reveals its gaps the moment it becomes a slide.
AI helped substantially here. Give it accurate inputs and guidelines and it cuts work time significantly while flagging points you had not considered.
At first I thought AI was making me safe. It produces the comfortable five-to-seven-out-of-ten answer. I complained that even when a bold move was needed, it kept returning safe, efficient options.
Over time I realized: AI was not giving me the safe answer. I was asking for it.
At some point, validating ideas with AI several times before a meeting became a habit. I wanted to be pre-validated before anyone could push back on me in the room. AI simply complied.
Bold ideas are, by construction, more likely to be argued with. So running them through AI repeatedly sands the bold parts away. AI was not blunt — I was getting blunter by running it again and again. It is the same problem identified in why Coca-Cola's AI Christmas ad failed: the value of what a prompt cannot make.
Everyone uses AI. Everyone uses the same sandbox to validate their thinking. The result is that ideas increasingly share the same grain.
The reason to hold a meeting is to spread out diverse possibilities and discuss them. An idea that has already passed through the AI filter is pre-refined thinking. Because it is not raw, it can be less compelling.
For the next pitch I am changing the order. I will not ask AI for the answer first.
I will spread my own thinking across several directions, build something close to a rough draft from that, and only then hand it to AI to develop. Getting the raw thought out before AI has a chance to censor me.
The contest is not decided by an AI-generated draft. It is decided by the one sentence that departs from that draft — and that sentence may only be available before you ask.
Adapted from an article provided through a partnership between Bruce and MobiInside.
Reviewing ideas and RFP fit without deference or preconception, which reduces team-wide confirmation bias, and exposing logical gaps when vague structure has to become slides.
Bold ideas invite pushback. Running them through AI repeatedly to preempt criticism filters out exactly the elements that made them bold.
Change the sequence. Spread your own thinking in several directions and build a rough draft first, then hand it to AI to develop — rather than asking AI for the answer up front.
To apply what you just read to your own site, start with a free audit of where things are now.
A strategist replies within 24 hours on business days.
LLM visibility depends on your whole digital footprint, not your website. Teams are still built by channel. Mordy Oberstein's uncomfortable advice: do not reach for an action plan yet.
Review content incrementality is the metric affiliate platforms cannot give you. Here is how one creator ends up paid four times over, and the six questions that separate acquisition from margin leakage.
An E-E-A-T checker was never automatable because there is no API for it. Here is how an AI coding agent read Google's quality guidance, crawled a page sample, and produced a scored audit — plus where it stops.