The AI Tender Trap: Why speed without soul loses work
Nathan Spruce
4/15/20266 min read
I want to start with a confession.
When I first watched an AI tool generate a 10 page tender response in under 10 seconds, my gut reaction was visceral and immediate:
I'd spent twenty years doing this the hard way. Pen. Paper. Whiteboard. Long sessions where the thinking happened slowly, out loud, in rooms with other people and the whiteboard wasn't just a surface to write on. It was an invitation. Anyone in the room could pick up that pen. Anyone could change something, add something, push back on something. That whiteboard was where co-design actually happened, before co-design was even a word people put in tenders.
And here was a machine doing in seconds what used to take days. But then I read what it produced and I felt it immediately ... the thing that was missing. Not a fact. Not a framework. Not a compliance item. Soul.
The kind of thing you can't name exactly but you absolutely feel when it's absent ... And evaluators feel it too. They just call it something different in the debrief.
The Trap Has a Shape
Growing up my old man was full of quips known in the family as "Bushyisms", one of them that has stayed with me for decades is this ...
Straight after he'd say it, I'd mimic it back with one of those screwed up kid faces that kids love to pull (chances are you've just relived it and pulled it just to see if you could). It would be years later that it would finally make sense. He meant that everything has a path, a framework, a method. You're never actually lost. You just haven't found the right way through yet.
I've spent twenty years proving him right in tender writing and one equation for AI is simple once you see it:
Speed + No Soul = A Faster Way to Lose. (For writing anyway)
Here's how the trap actually works.
AI makes you faster. Faster means you can take on more. More work means more pressure. More pressure means you lean on AI harder. Leaning harder means the soul gets thinner and a thinner soul means lower win rates ... which creates more pressure to work faster. It's a vicious circle dressed up as an efficiency gain.
If not yourself, you know someone it's happened to. Genuinely talented. Experienced. But chronically time-poor in a way that breaks people. AI gives what feels like breathing room; fast drafts, quick structures, responses that ticked every visible box.
What it actually gave was the illusion of progress while quietly eroding the thing that makes stuff worth reading in the first place.
What AI cannot do (and will never do)
Let me be specific, because this matters.
AI can match evaluation criteria. It can structure a response logically. It can draft compliantly and quickly and without obvious errors ..... when you prompt it well.
But it cannot pick up a whiteboard marker and change something because it felt wrong in the room.
It cannot sit in a community co-design session and feel the shift when someone says "the last lot promised us that too" and know that line needs to be in the response somehow, without making it exploitative or performative.
It cannot carry twenty years of knowing what a winning room feels like.
In HBDI terms and I use this framework constantly, AI lives almost entirely in the Blue and Green quadrants. Facts, logic, process, structure. It can approximate Yellow if you push it. But the Red quadrant the relational, empathetic, human layer is not something it can generate. It can only imitate it. Experienced evaluators can feel the difference between genuine relational insight and a very good imitation of one.
I've seen AI-heavy responses that are technically strong but winning bids read like it was written by people who had walked the job site at 6am, listened to what the community actually said and had the courage to put that truth on the page.
That's not luck. That's the Red quadrant doing its job.
The 50% threshold and why it's already here
Here's something I know directly, not from an article, not from a conference panel. Internal AI detection is here.It scans for things like:
Repetitive sentence structures
Generic phrasing patterns with no specific evidence
Over-reliance on corporate-speak
Absence of verifiable, particular detail
When I was chatting about this with a colleague, he casually threw in that, if requested by the client, his work screened submissions for over use of AI. Submissions that score over 50% for AI-generated feel are automatically filtered before a human evaluator sees them.
Not reviewed sceptically. Not flagged for discussion. Removed from the first round entirely.
The reason they built it wasn't hostility toward AI. It was the sheer volume of submissions that had become indistinguishable from each other. AI raised the bar and every response became competent ... but it also collapsed the ceiling. Nothing stood out. Nothing felt like it came from people who genuinely wanted the work.
They built the filter to find the humans in the pile.
The system that actually works: Human-first, AI as tool
My dad's equation still holds. There is a path through this. Here's the one that works.
Step 1: Start at the whiteboard. Literally.
Before AI touches anything, get your team in a room. Put the brief on the board. Write the win themes. Map the community stories. Identify the relational insight only your team has; the thing you know because you were there, because you've done this work in this place with these people.
Invite people to pick up the pen. Let something change. That's not inefficiency, that's co-design. It's what makes the eventual response feel like it came from humans.
If you have an HBDI team profile, run a quick debrief here. Find the Red quadrant gaps and name them before anyone opens a laptop.
Step 2: Use AI for the first draft. Nothing more.
Feed it your whiteboard notes. Feed it the brief. Feed it specific evidence placeholders where you don't yet have the detail. Ask it to draft in plain language. Then step back. Read what it gives you and treat it exactly as it is: a starting point that has none of the soul you're about to put into it.
Step 3: Inject the human layer. This is the work.
Read every section aloud. Ask three questions:
Does this sound like us?
Where's the lived experience?
What's the one specific moment, the conversation, the site visit, the community story that only you could have written?
This is the step that takes time. It's also the step that wins tenders.
Add the reader cameo. Not "we engaged with stakeholders" but the actual texture of what that engagement revealed and why it changed your approach. This is the HOW and for those that have worked with me before you know how I bang on about this (It's also coming up as a separate article). Evaluators have read a thousand versions of the first sentence. Almost none of the second.
Step 4: Let AI edit, not author.
Run the human draft back through AI with one specific instruction: tighten the flow, remove corporate filler, check compliance language but do not touch a single human story or relational insight.
This is AI doing what it's genuinely good at. Cleaning. Not creating.
Step 5: The human proof.
Then, more importantly, have someone read it cold as if they were the evaluator. Not checking boxes. Checking conviction. Asking themselves: do I believe these people should win this work? And if it were my money, would I give it to them?
If yes, submit with confidence. If not, back to the step 3 and quite probably the whiteboard.
The thing worth saying plainly
I'm not anti-AI. I use it. I'll keep using it. I started this work with a pen, a whiteboard and the belief that every brief has a human being behind it who deserves a human response. Twenty years of that doesn't disappear because a tool got faster.
The whiteboard is still where the best thinking happens. Not because it's old-fashioned — but because it puts a pen in someone else's hand and says your thinking matters here too. It also allows you to draw the answer and connect the dots. That's co-design. That's the Red quadrant. That's the thing AI will never replicate.
In 2026, with procurement teams actively filtering for the presence of human conviction, it's also the thing that wins.
My dad was right. There's an equation for every occasion. This is today's
Your experience + your stories + AI as a tool = work worth reading.
Find that equation. Use it every time.
More Bushyims to follow ...