Work our team has delivered · CONSTRUCTION · SUBCONTRACTOR
A tender agent that cut total tender response time by 60% for a construction subcontractor.
A tier-2 subcontractor was turning down about two in five tenders because its estimators ran out of hours. Our team built an agent that drafts a first response in under 30 minutes, from 15 years of past wins. A senior estimator still reviews and owns every word that goes out.
↳ Some of this work was delivered by Reenva's founder and team before Reenva was formed. Client name withheld; some commercial figures shown as bands.
/01 · What changed
60% reduction in total tender response time, with drafts in the firm’s own winning words.
Estimators no longer rebuild the same capability statements from scratch. The agent drafts in the firm’s own proven words. A senior estimator reviews it, sharpens it and owns the price.
60%
Reduction in total tender response time
21 days
From kickoff to the first agent-drafted tender sent
<30min
For the agent to assemble a first draft from past winning bids
9wk
From kickoff to the agent live across the tender pipeline
/02 · The problem the team was asked to fix
Each tender came with a deadline, and building a response took a senior estimator two or three days. When no estimator had that time, the firm didn't respond badly. It didn't respond at all.
About two in five tenders were declined this way. The reason was capacity, not price or fit.
The firm already had what it needed to answer them. Fifteen years of winning submissions sat in SharePoint: capability statements, methods, and safety and quality sections. But nobody could find the right one in time, so every tender started near zero.
Why a generic tender writer wouldn't fix it
A subcontractor wins on specifics: named certifications, real project references and a method that matches how it runs a site. A generic AI writer strips those out, and the people scoring the tender notice. The winning language was the firm's own; it just had to be easy to find.
/03 · How the team worked
The team learned how the firm wins tenders before building anything.
For two weeks, the team’s engineer sat with the estimators through live bids. The aim was to learn what 15 years of winning sounded like.
- 01 · Beside the estimators
The team sat with the estimators through live tenders.
The team’s engineer watched a senior estimator rebuild the same capability statement from memory, again. Two jobs were declined late on a Friday because no hours were left. The problem wasn’t skill, but where that skill was spent.
- 02 · Show the cost
A number on the bids the firm never sent.
The team rebuilt the declined pipeline and put a value on it. The costly loss was tenders never started, not tenders lost on price. Once leaders saw that figure beside their win rate, the brief became “stop us leaving this on the table.”
- 03 · Re-engineer the process
Redesign the response, then hand the first draft to software.
The tender process was redesigned so the first draft comes from the firm’s own past wins. Estimators now spend their hours on pricing, risk and final edits. The agent follows the new process; the estimator owns the result.
The agent never sends a tender. A senior estimator still owns every word that goes out. They get back the two to three days a first draft used to take.
/04 · The cost, in numbers
The costly tenders weren’t the ones the firm lost. They were the ones it never sent.
Most win-rate talk is about tenders submitted. The bigger cost was tenders declined at the door, because estimators had no hours left that week.
↳ The team rebuilt this from 15 years of past submissions: bids won, lost and never started.
Reconstruction · 15 years of past submissions
- Live tenders received each year
- 50–60
- Share previously declined for lack of capacity
- ~40%
- Estimator time to build a first draft by hand
- 2–3 days
- Typical contract value of a tendered job
- $250k–$1.2M
- Win rate on tenders actually submitted
- ~1 in 5
Source: anonymised tender register, 15 years of the firm’s past submissions in SharePoint, and the Dynamics 365 pipeline. Figures shown as bands; client details withheld.
/05 · What the team did
Measure the cost first, fix the process, then build.
- /01
Diagnose.
Two weeks with the estimators during live tenders. The team mapped where the hours went and which sections were rewritten every time. It also put a value on the work declined for lack of capacity.
Delivered: a map of tender effort, the value of declined work and an agreed first build
- /02
Prepare the data.
Fifteen years of past submissions sat in SharePoint folders, named by whoever saved them. The team turned them into one searchable library of the firm’s winning language. Each document was tagged, so the right clause can be found for the right job.
Delivered: a searchable library of past submissions, linked to the Dynamics 365 pipeline
- /03
Build.
The tender agent reads each incoming request for tender and finds matching methods and clauses from past wins. It assembles a first draft in the firm’s own words in under 30 minutes. The estimator reviews, prices and signs: a person always owns the last step.
Delivered: the agent live on the tender pipeline, with first drafts in under 30 minutes
- /04
Run and improve.
Every submitted tender is added back to the library. Win and loss reasons are recorded in Dynamics 365, so the firm knows why it wins, not just how often. Each new win gives the agent better material to draw on.
Delivered: win and loss reasons recorded, plus a dashboard of tender capacity
/06 · Technology the team used
Data layer
- SharePoint documents organised and indexed
- 15 years of past submissions made searchable
- Dynamics 365 tender pipeline connected
- Win and loss reasons recorded
What the team built
- Tender agent on Azure OpenAI that searches past winning bids
- Reads each request for tender and lists its requirements
- Finds matching clauses and methods from won bids
- Assembles a first draft in the firm’s own words
Reporting layer
- Power BI tender pipeline dashboard
- Win rate by sector, estimator and value band
- Response-time tracking
- Capacity against demand
/07 · What changed for each team
Sales floor
Estimators stopped retyping the sections that never change. They now spend their hours on pricing and risk, the judgement that wins work. The business development team now answers tenders it used to decline for lack of time.
Operations
The tender register is no longer a folder structure and someone’s memory. Every request, response and result is recorded and searchable, with a reason on every loss. Writing time is no longer rationed to whoever is free.
Board
The board can now see the full tender funnel: received, answered and won. It can also see the value once lost at the door. Response capacity is a tracked number, not an excuse for a declined bid.
“We used to decide which tenders to even attempt based on who had a free afternoon. Now the first draft is waiting when we sit down. The judgement is still ours — we just get to spend it on the part that wins.”
/09 · Your next step
Facing the same problem? Find out what fixing it would take.
If there is a fit, a two-week fixed-fee review follows, ending in a clear go or no-go.