A Week On Every Tender Is Where Your Margin Goes

Tender prep at one of the trades businesses we work with used to take a full working week. It now takes about three hours.

Same team. Same type of work. Same clients reading the submissions on the other end. They stopped asking their most expensive people to do the clerical half of the job, and gave that half to AI.

I think tendering is the clearest margin leak in trades and construction. Everyone in the business can feel it. Almost nobody puts a number on it.

Put A Number On Your Last Tender

Think about the last tender you submitted and count the hours honestly. Reading a 100-page request for tender, twice. Hunting through it for the mandatory requirements buried in an appendix. Opening the last bid, copying the safety management plan and changing the names. Rewriting the capability statement because the template asks the question slightly differently. Chasing three CVs, a licence renewal and a certificate of currency. Formatting. Fixing the formatting.

Somewhere in there is the actual estimating. The price, the program, the risk. It’s a smaller slice than most owners expect.

Now run some rough maths. Say your estimator or operations manager costs the business $150 an hour once super and overheads are in. Forty hours on a tender is $6,000. Bid twenty a year and that’s $120,000 of senior time before you’ve won a thing. If you win one in four, three of those weeks bought you nothing but practice.

Your numbers will be different. Work them out anyway, because that figure is the size of the leak.

The Clerical Half And The Judgement Half

Every tender splits into two piles of work.

The clerical pile is reading, extracting, matching and assembling. Which requirements are mandatory. What the evaluation criteria are and how they’re weighted. Which past project best answers “demonstrate experience on similar works”. What we wrote about traffic management last time. Pulling it together inside the word limits.

The judgement pile is everything only an experienced person can decide. Should we bid at all? What’s the price? What risk are we carrying in those liquidated damages? Which supervisor do we name, and are they actually free in March? What are we prepared to commit to?

AI has become very good at the first pile. It will read a 120-page document in a minute and hand you every mandatory requirement with a page reference. It will find the answer you wrote for a similar job eighteen months ago. It will draft a methodology from your past winning submissions, in the language your business already uses.

It has no business near the second pile. It doesn’t know that the client’s project manager fights every variation, or that your best crew is booked solid until after Easter. Your estimator knows. That’s what you’re paying them for.

What The Three-Hour Version Looks Like

The workflow is less exciting than people hope. Five steps.

Build a bid library. Past tenders, capability statements, CVs, licences, insurances, safety documents, project case studies and photos, all in one organised place. This is the boring bit and it’s most of the setup. AI can only draw on what you give it, which I cover in more depth in getting your data ready for AI. A messy shared drive gives you a messy draft.

Run a requirements pass. AI reads the request documents and produces a compliance checklist, with page numbers, plus a list of anything unusual. Insurance levels above what you carry. Tight timeframes. A clause that shifts design risk onto you.

Make the go or no-go call. A person does this, with the checklist in front of them. Fifteen minutes, maybe thirty. A lot of the value lives here, because a clear summary makes it much easier to walk away from the wrong job early.

Assemble the first draft. AI pulls the relevant material from the library, matched to the evaluation criteria and the word limits, and flags every gap where it needs something it doesn’t have.

Expert review. Your estimator spends their time on price, program, risk and the job-specific detail that makes the methodology believable. That’s where the three hours go.

Another business we worked with prepared a $4 million tender with about three hours of work this way. The scale of the job doesn’t change the shape of the process much. Bigger tenders mostly mean a longer checklist and more careful review.

Where It Goes Wrong

I’ve seen this go sideways in a few predictable ways.

Someone pastes the full tender pack into a free chatbot on their phone. Tender documents often carry confidentiality conditions, and consumer AI accounts may keep what you give them. Use business accounts with the data controls switched on, or tools the business controls. If you don’t know what your team is already doing here, shadow AI is worth reading before your next bid.

Someone trusts the compliance checklist without checking it. These tools are probability engines. They are usually right and occasionally, confidently wrong, and a single missed mandatory requirement can make a bid non-conforming. Every mandatory item gets checked against the source page by a person. That’s why the page references matter.

Someone lets the draft go out generic. Evaluators read stacks of these and they can spot a template from the first paragraph. What wins is specific: the named supervisor, the real project down the road, the actual crew size and how you’ll keep the site safe with a school next door. AI gets you to a solid draft fast. The specifics are your job, and skipping them produces the polished nothing I wrote about in AI workslop.

And sometimes the whole thing gets built inside one enthusiastic person’s laptop. When they go on leave, tendering goes back to a week. Document the workflow and make sure two people can run it.

The Tenders You Never Sent

The hours are only part of the cost. The bigger one is harder to see.

It’s the work you passed on because nobody had a spare week. The council job you’d have loved, which closed while everyone was flat out on site. The repeat client who asked you to bid and got a polite no.

When a tender costs three hours, the decision changes. You can bid on the work you actually want and say no to the jobs you were only chasing because the pipeline looked thin. Your win rate starts to reflect your choices instead of your capacity.

Your estimator gets most of a week back, every time. What they do with that week is where the growth comes from.

Frequently Asked Questions