Construction and Trades

The site already told you. Write it down once.

Quotes, site records and handover documents live in email because the work happens somewhere the systems are not. We build capture that fits a site and a job record that everything else attaches to.

  • Site capture and daily records
  • Quote to job on one record
  • Cost, variation and valuation
  • Client communication kept human

The record gets written after the fact, if it gets written

Every one of these is a place where somebody knew something and it never reached a system that could act on it. Six patterns turn up in almost every contracting business we map.

The day gets written up on Friday, from photographs

Nobody filled in the daily report on Tuesday because they were on site. By Friday the person writing it up is reconstructing four days from camera timestamps and memory.

A site record that is thinnest exactly where a dispute will land.

The quote, the order and the invoice are three separate documents

What was priced, what was ordered and what was billed each live in their own file, and the differences between them surface at the valuation rather than at the point somebody could have asked about it.

Variations you delivered and never charged for.

A slipped delivery is recorded and nothing works out what it moves

The supplier emailed a new date, somebody read it, and the three work items waiting on it kept their old dates because updating them meant walking the programme by hand.

A completion date that moves without anyone deciding it should.

Handover is assembled at the end, out of an inbox

Certificates, warranties, test results and as-built information were all produced during the job and none of them were filed against anything. Somebody now spends three weeks finding them again.

Retention held while the paperwork is rebuilt.

Subcontractor status is whatever the last phone call said

Who is on site, what they have completed, what they are waiting on and what they have claimed for lives across a programme, a WhatsApp thread and somebody knowing.

Two trades booked into the same space on the same morning.

The same information gets asked for four times

FMI and PlanGrid surveyed nearly 600 construction professionals and found around 14 hours a week going to looking for project information, resolving conflicts and dealing with mistakes and rework.

A third of the week spent finding out what already happened.

Give the system the writing up and keep the commercial calls

Controlled delegation is the whole design, and everything below this band is an application of it. The work of recording and working out goes to the system. What the client is told, what gets claimed and which job a cost belongs to goes to a person.

  • 01

    The work happens away from the system

    A delivery arrives, a trade finishes a section, a supplier moves a date, an inspection passes. All of it is known by somebody standing on a site, and none of it is anywhere yet.

  • 02

    The system takes the writing up and the working out

    Turning what was said into a report, filing photographs against the work they show, matching a delivery to its order, and walking the programme to see what a slipped date moves. None of it is judgment.

  • 03

    People take the commercial calls

    What the client is told, what gets claimed, which job a cost belongs to and whether a delay is recoverable. These reach a named person with the working already done.

See it run

Forty seconds of talking becomes the day record

A voice note, three photos and a delivery slip go in. The daily log, the programme consequence and the cost record come out, and anything the client would see waits for a person.

Site to Office
  1. Input
  2. Context
  3. AI work
  4. Action
  5. Approval
  6. Recorded

Connected systems

  • Field app Voice, photos, slips updated
  • Project management Daily logs and issues updated
  • Schedule Programme and dependencies updated
  • Cost system Committed and actual updated
  • Documents Photos and paperwork updated
  • Client comms Notices and variations updated

Site update Kingsmead Phase 2 Block C, Thursday

  1. Voice note received 38 seconds, from the site foreman
  2. 3 photos and 1 slip Uploaded on site, geotagged
  3. Project identified From the geotag and the foreman record
  4. Programme pulled Current activities for Block C
  5. Voice note transcribed Trade terms and two supplier names recognised
  6. Work matched to activities Second fix electrical, level 3
  7. Photos attached to work Each one to the activity it evidences
  8. Delivery slip read 12 items, matched to the purchase order
  9. Daily log written Progress, labour, weather, deliveries
  10. Progress updated Second fix at 80%, on programme
  11. Delivery recorded Against the purchase order, not re-keyed
  12. Photos filed Named, dated and tied to activities
  13. Held for the project manager A person signs the day off, in 30 seconds
  14. Run recorded Transcript kept alongside the record it produced

Result

Daily log written The record exists because making it cost the foreman forty seconds.
  • Voice note 38 seconds
  • Photos filed 3
  • Activities updated 4
  • Second fix 80%
  • Programme On plan
  • Typing on site None

Completed automatically

  • Voice note transcribed
  • Daily log written
  • Programme progress updated
  • Delivery matched to the purchase order
  • Three photos filed against activities

Requires a person

Signing off the day record.

Project manager

Sign offAmend the logQuery foreman
View audit trail
  • 16:41 Voice note received from the field app. Geotag matched to Kingsmead Phase 2.
  • 16:41 Transcript produced and retained with the log.
  • 16:42 Delivery slip matched to purchase order 4471. 12 of 12 items.
  • 16:42 Programme progress written. Second fix electrical 80%.
  • 16:42 Held for project manager sign-off.
5 actions completed 4 systems updated 1 exception 1 approval

This demo uses fictional data. The workflow gets configured around your systems. Build a workflow like this

Four workflows, and the order they usually go in

Site capture usually goes first because the site feels it immediately and because it starts filling the record everything else needs. Each card names what the system handles and where a person stays in control.

Operations

Site capture to the day record

A voice note becomes the report, the photos and the issue

The foreman talks for forty seconds. The system writes the daily report on your own format, attaches the photographs to the work items they show, matches deliveries to their purchase orders, and raises anything that moves a date as an issue with the impact already calculated.

  • Capture that fits the site talking, not eleven fields on a phone in the rain
  • Photos filed against the work not into a folder nobody opens again
  • Programme impact worked out what a two day slip actually moves
How a workflow gets built

Data: site capture, the programme and your purchase orders. Human: the project manager approves anything the client sees.

Revenue

Enquiry to quote to job, on one record

The same reference from the first call to the final account

Enquiries, site visits, quotes, revisions and won work on one record, so the firm can say what it quoted, what it won, at what margin and why it lost the rest. The estimating half of a construction business is usually the least systematised part of it and the one closest to the money.

  • One record per opportunity every revision of the quote against it
  • Win and loss with reasons which client types and which work you should stop bidding
  • Quote to job without retyping the won quote becomes the job it was priced as
How the record gets built

Data: enquiries, quotes and job records. Human: an estimator prices it and a director approves the terms.

Operations

Cost, variation and valuation

What was priced against what is actually being spent

Purchase orders, deliveries, labour and subcontractor applications landing on the job they belong to as they happen, so the cost position is current rather than assembled monthly. Variations get raised as variations at the point the instruction arrives.

  • Costs land on the job at the point they are committed, not at month end
  • Variations raised when instructed rather than remembered at the final account
  • Applications built from the record with the evidence already attached
How a workflow gets built

Data: orders, deliveries, labour and applications. Human: the commercial manager owns allocation and valuation.

Insight, later

Which jobs actually made money

Margin by job type, client and estimator

Realised margin by work type, by client and by who priced it, with the overruns traced to whether it was the estimate, the programme, the subcontractor or the client that caused them. This one comes last because it stands on the record the other three build.

  • Margin after the final account not the margin on the day you won it
  • Overrun traced to a cause estimate, programme, supply or client instruction
  • Client profitability the ones worth keeping and the ones worth pricing higher
How the data layer gets built

Data: your own job costing and delivery records. Human: directors decide what the firm stops bidding for.

The programme is the part that has to be real before any of this is interesting. A system can only tell you what a two day slip moves if something says what waits on what. Most contracting businesses have that in a planning tool that nobody has updated since week three, and getting it back to being true is usually part of the first engagement rather than an assumption underneath it.

Six decisions taken before anything is built

These get agreed with whoever runs the sites and whoever owns the commercial position before a build starts. Between them they decide what the workflow is allowed to be.

What gets captured on site, and by whom
Which people can record against a job, what they can change afterwards, and whether the original audio, photograph or note is retained alongside what was written from it. On a contested job the original is the evidence.
Cost allocation rules
Which costs a workflow may post automatically and which have to be allocated by a person. A cost posted to the wrong job is quiet at the time and expensive at the valuation, which is why the demo above refuses to guess between two projects.
Client communication Kept human
Anything that tells a client a date, a cost or a scope has changed is drafted by the system and released by a named person. There is no threshold under which this becomes automatic.
Health and safety records
What the workflow may record about incidents, inspections and near misses, how long it is kept and who can see it. This is usually governed by something outside the business, and the build follows that rather than the other way round.
Audit trail
What was captured, when, by whom, what the system wrote from it and what a person changed afterwards. Enough that the site record can be relied on in a dispute rather than argued about.
Escalation path
What happens when a delivery cannot be matched, a programme dependency is missing, or a capture names something the job does not have. A workflow with no escalation path escalates to whoever notices.

What counts as working. Accepted output against your own baseline is the proof. The baseline gets taken before the build exists, and afterwards the same work is measured against it with the review burden counted openly, because a system that produces plenty and has to be corrected twice has saved nothing. How we measure AI work sets out the method.

One workflow, proved, before anything bigger

Roughly eight to twelve weeks from the first session to a workflow running against a baseline you set, on one live job rather than across the whole business.

Map the work
Two to three weeks. How enquiries, site delivery and cost actually run, where the time and the leakage sit, and the candidates scored by value and readiness so the first move is an evidenced choice. This is FusionMap.
Build the record
Five to nineteen weeks depending on scope. The job record that costs, programme and capture all attach to. Most construction businesses have three systems that each hold part of a job and none that holds the job.
Build one workflow
Four to six weeks. Usually site capture, because it is the one the site feels on day one and the one that starts filling the record the rest of the work needs.
Train and operate
Ongoing. Into daily use with the site trained to run it, reviewed against the baseline, then the next workflow if the first one earned it.

Where this has already been built

The approach on this page comes out of these engagements. Two are businesses that move physical goods and run on process, and two are the design patterns the workflows above are built on.

Brand, growth and process for a construction materials supplier

Argent supplies furnishing and construction materials and was expanding into new markets with the operation held together by people. Brand strategy, growth marketing and process improvement, delivered together because the demand would have broken the operation on its own.

New markets expansion supported by the process work Brand and ops delivered as one engagement
Read the case study

A digital adoption plan for a manufacturer, accepted by ISED

Inventory and operations mapped across every department for a specialty manufacturer, with the plan assessed and accepted by a funding body. The same discipline a construction business needs before it buys anything: know what the operation does before choosing what runs it.

Every department mapped, not just the obvious ones Accepted by ISED as a funded plan
Read the case study

Research that arrives before the call

Territory research a team used to do by hand now lands every morning without anyone triggering it. The pattern behind the enquiry workflow above: assembling that a person would do badly at nine in the evening, done before the day starts.

550+ districts researched per territory 3 agents kept running and updated
Read the case study

A payments platform where the controls were the product

An escrow exchange built from nothing, for people sending money to counterparties they have no reason to trust. Designed, built and deployed, with the controls as the feature rather than as a wrapper around it.

Full stack designed, built and deployed Escrow funds held until both sides are satisfied
Read the case study

Begine Fusion is a Zoho Authorized Partner working across Canada, the United States, the United Kingdom and Nigeria, with a team on the ground in Nigeria since 2013. Every engagement is published in full, including what was measured and what was a design target. The full set is on our work.

Whether this is the right conversation yet

The work pays back when the business is running enough jobs at once that no one person holds them all. If that is not where you are, the right column says what to do instead.

This fits if

  • More than a handful of jobs run at once and the daily record is written up after the fact
  • You have found a cost on the wrong job, or a variation you delivered and never charged for
  • A supplier date moved and the programme found out about it weeks later
  • Handover packs get assembled at the end out of an inbox
  • Somebody can own the job record, because everything on this page attaches to it

Start somewhere else if

  • You do not yet know which jobs lose money. Go to FusionMap and find out before commissioning a build.
  • The programme has not been updated since the job started. That gets fixed first, and it is usually cheaper than any workflow on this page.
  • You are looking for estimating software rather than a system. Buy the software, then come back when it stops being the problem.
  • Nobody is looking after the systems you already have. Go to Managed Operations.

Start where the problem is

The same five offers every other client buys, scoped to a business whose work happens away from its systems. Find the line that sounds like yours.

The jobs are busy, the margin is unpredictable, and nobody can say which jobs did it

FusionMap

Quotes, jobs, costs and site records live in four places that do not agree

Systems Build

One workflow is defined, agreed, and ready to be built properly

FusionBuild

People are already using AI on site reports and client correspondence, unofficially

FusionGuard

The systems went in, they work, and nobody has looked after them since

Managed Operations

Your people have to supervise what a system produces, not just operate it

AI Systems Mastery

No price is quoted on this page. A sector is not an offer and does not have a price of its own. What a business here pays follows which of those six lines is the real one and how much of the job record already exists. Working out which line is yours is what the first call is for.

Asked on every call with a contractor

The seven that decide whether a build is worth starting, answered before the proposal rather than inside it.

Our site teams will not use another app. Is this another app?

The capture side is built around what already happens: talking, photographing and forwarding. If a workflow needs a foreman to open something and complete a form, we have designed it wrong and it will be abandoned inside a month. The systems that need looking at are the ones in the office, and site should feel like less work rather than more.

What about the sites with no signal?

Capture queues on the device and syncs when there is a connection, which is the normal arrangement and worth confirming rather than assuming. What matters more is that a voice note recorded in a basement at 14:40 still carries 14:40 when it lands, so the record says when the thing happened rather than when the phone found a bar of signal.

Does the system talk to the client?

No. It drafts, and a named person releases. Telling a client that a date has moved is a commercial act with a contract behind it, and a two day slip is often recoverable, which is a judgment about the job rather than a fact about the programme. The demo above stops at exactly this point on purpose.

We already run project management software. Do we replace it?

Usually not. The common finding is that the programme is in one system, the costs are in another, and site capture is in a third that is really a phone. That is a connection problem. If the mapping finds a system that genuinely cannot carry the work, we will say so with the reasoning rather than a recommendation.

Can it work out delay and disruption for a claim?

It can show what moved, when, on what evidence, with the source document attached to each date change. Whether that constitutes a claim, and how it is put, is for your commercial team and whoever advises them. A system that produced a claim on its own would be producing an argument, and an argument nobody can trace is worth nothing when it is tested.

How do we know it worked?

A baseline is taken before the build exists: how long the day record takes, how much gets reconstructed later, how much cost lands on the wrong job, and how late a programme change is currently noticed. Afterwards the same work is measured against it, with the review burden counted openly.

How long before anything is running?

Site capture is usually four to six weeks once the job record exists. If it does not exist, that comes first and runs five to nineteen weeks depending on scope. Ahead of both sits the mapping, which is two to three weeks and produces a decision you can act on whether or not we do the build.

Bring the job where the paperwork cost more than the work

Thirty minutes with a director, whoever runs the sites and whoever owns the commercial position is enough to name the first workflow, the record it needs underneath it, and the baseline it gets measured against.