Built Logic

Reading an M&E spec went from 6 hours to 2.5

At an M&E contractor in Greater Manchester with 110 staff, the estimators used to spend about 6 hours reading each tender’s spec and prelims. It takes about 2.5 now, and work that isn’t in the price gets qualified at tender, not carried on site.

Key results

  • 6 → 2.5 hoursReading the spec and prelims
  • 90 → 30 minsWriting the qualifications letter
  • 6 to 8 → 2Hours checking a subcontract at award
The business
M&E contractor
Service
Spec Checking
Live in
5 weeks
Paid back in
Approx 7 weeks
Size
110 staff

Background

They price about 85 tenders a year. A tender pack can run to 1,300 pages, and they get about 8 working days to price it, often with 3 other tenders on at the same time. With that workload, every hour on the spec was an hour off the take-off.

The challenge

Reading the spec and prelims took about 6 hours a tender. Writing the qualifications letter from scratch took another 1.5, so over a year it came to 640 hours.

A 540-page M&E spec can move work onto the M&E package on any page, and every consultant words it differently. Fire stopping, for one, is by others on some jobs and in the M&E package on the next.

At award, the commercial manager checked 14 subcontracts last year, at 6 to 8 hours each, against the tender version.

640 hoursa year reading specs and writing qualifications letters

The solution

A plain bit of code keeps every page number and won’t let a flag through without one. AI reads the spec and the subcontract against the commercial manager’s 31 red lines, because every main contractor words its subcontract differently.

What we built

  • Checks the subcontract. It reads it clause by clause against the 31 red lines, and every flag shows its page.
  • Finds who does what. It checks 120 M&E scope items against the spec, such as whether fire stopping is in the package or by others, each with its page.
  • Drafts the qualifications letter. It uses the flags the estimator keeps, in their usual wording. The commercial manager approves them in Teams.
  • Compares at award. It checks the subcontract that arrives against the tender version and flags anything new.

Tech stack

It runs on their own Microsoft 365 and Azure accounts. It works alongside the tools they already use.

SharePoint Teams Word Power Automate Azure AI Document Intelligence Azure OpenAI Copilot Chat NBS Chorus Bluebeam Revu Claude, on Azure

The results

  1. 01

    Work that isn’t in the price, qualified at tender

    3 scope items in the first 21 tenders: fire stopping, a connection to a district heat network, and attendance on the client’s commissioning engineer. Each one is in the qualifications letter with its page, so it’s priced, or excluded and agreed, before anyone signs.

  2. 02

    Retention held at 3%

    On one of the first jobs, the subcontract sent at award had retention at 5%, not the 3% in the tender. They wouldn’t sign until it went back to 3%, and it did.

  3. 03

    21 tenders checked against the red lines

    Checked against the red lines at tender, with page references. About 1 in 6 flags is cleared on a second look.

  4. 04

    2 red lines changed in the first month

    The commercial manager owns the list, so when they change a red line, the check changes with it.

Find what nobody’s pricing in your business

Answer a few questions about your business and you’ll get your top three priorities, the changes that would put the most margin or hours back, and where to start. It costs nothing.