The AI executive layer for general contractors
Updated in real time, across every department. Not at closeout.
dideh connects every department, from the field to the back office, into one governed model of your company. It sits on top of the tools you already use, with nothing to implement or migrate, and shows you what’s true right now so you can see what’s coming next. Your leadership team works from the same facts, and you see the whole business as clearly as you once saw every job.
Ecosystem
dideh reads from your ERP, accounting, project, scheduling, field, CRM, payroll and document systems. Nothing to migrate, and nothing changes in the tools your teams use every day.
Read-only connections
Data not ready to connect? Our data foundation team cleans it first.
Ask dideh
Ask in plain English. dideh works the answer out from certified measures, your systems and the documents behind them, shows each step it took, and tells you what it means and what to ask next. Every answer is governed for the seat that asks.
What it draws on
Question: Provide win rate by region.
dideh took 4 governed steps: Reasoning, 20 words; Certified measure, win_rate — Win rate (TTM) · certified · owner President · Estimating · higher is better; Certified measure, win_rate_dollars — Win rate by value (TTM) · certified · owner President · Estimating · higher is better; Semantic SQL, 5 rows · from bid · 0.4 ms.
Over the last 12 months the company won 33.6% of the bids that reached a decision. Mountain is highest at 53.8% and Central lowest at 24.0%.
| Region | Won | Lost | Win rate | By value | Won value |
|---|---|---|---|---|---|
| Mountain | 7 | 6 | 53.8% | 48.9% | $71.4M |
| Southeast | 9 | 16 | 36.0% | 30.2% | $84.7M |
| Northeast | 5 | 10 | 33.3% | 38.6% | $62.5M |
| Pacific | 9 | 20 | 31.0% | 27.1% | $108.3M |
| Central | 6 | 19 | 24.0% | 18.7% | $49.6M |
From the certified measures win_rate and win_rate_dollars (owner: President). Both count only bids won or lost after Sep 30, 2025, so shelved and open bids are left out.
What it means
Suggested next questions: Why is Central losing bids: price, competitors or owner relationships? Which competitors beat us most in Central and Pacific? Which architects bring us our highest win rates?
Sample company, modeled data
What it draws on
The problem
Cost, billing and schedule drift apart during the month, and the explanation lives in someone’s head.
The problem shows up after close, once cost, billing and schedule have already diverged.
The estimate, the commitments, the field and the cash forecast each tell a different version of the same job.
Critical context sits with the one project manager, controller or executive who knows where to look.
Cash scenario
Set your revenue, then pick what goes wrong. The arithmetic shows how much less cash you would end the year with, and how much pay-when-paid gives back.
On $400M of revenue, you would end the year with $15.2M less cash, about 2 weeks of billing. Owners paying 30 days later hold up $32.9M, and pay-when-paid passes $21.7M of it to your subcontractors. A fade of one point costs $4.0M.
Arithmetic on your revenue, not a forecast. Slower pay: $400M ÷ 365 × 30 days = $32.9M, of which pay-when-paid passes about two thirds to subcontractors. Fade: 1% × $400M = $4.0M.
dideh runs this week by week on your own jobs, owners and subcontracts.
See the cash forecastHow it works
dideh builds a context layer over the systems you already run: your business described in its own terms, with one set of rules for every person and every AI specialist who reads it.
Connect
Job cost, payables and receivables from your ERP; budgets, commitments and pay applications from project management; schedules, field logs, CRM, payroll and documents. Read-only, with every field traced to its source.
Model
Jobs, cost lines, change orders, pay applications, vendors, owners, architects, bids and people become one model, with the relationships between them. Backlog, fade, DSO and over/under billing are defined once, assigned to an accountable executive and tied out to your books.
Govern
Who can see a row, which values are masked, who may approve what and up to how much, and an audit trail of every query, answer and action. Each specialist is a seat too, with only the access its mandate needs.
Act
Each leader opens a brief for their seat and can ask anything in plain English. Specialists watch the measures they own, propose the next move with evidence and a draft, and route each decision to the person accountable for it.
A look inside, on the sample company
Certified measure
CertifiedEstimated margin today minus the margin carried in the bid, for every open job.
−0.6 pts−$4.2M on the open book
Tie-outs
A measure is certified only when its tie-outs pass, so a number in a brief, an answer or a proposal is the number in your books.
Evidence
Status reports, OAC minutes, RFIs, delay notices and change-order narratives are linked to the job, the owner, the architect and the vendors, so an answer can say why and show where it read it.
OAC minutes · Meeting 14 · Sep 18
Owner approved the revised stair finish; the cost isn’t in the forecast yet.
Delay notice · Aug 29
Owner-directed finish change, 11 days claimed.
Change-order narrative · CO 23
$186K waiting on the owner’s signature.
The model
dideh models the chain your company already runs, so every number traces back to the job, the commitment and the people behind it.
Bids, win rate and estimated margin
Win rate
Contract, terms and baseline schedule
Contract value
Buyout, commitments and vendors
Bought out
Earned, billed and retained
Over / under billed
Collections, payables and liquidity
13-week cash
Pay-when-paid from owner to every subcontractor and supplier, retainage on both sides, and fade traced back to the estimate.
The platform
One model, six views. Each is governed for the seat that opens it.
The Today screen for a chief executive: revenue, gross margin, cash and backlog, with the decisions and alerts for them: a pursuit to decide, a warm introduction, a relationship at risk.
Sample company with modeled data.
A decision routed to the CFO: revise a pay application to bill $412K of earned work, with the evidence and the reason it needs CFO approval.
Sample company with modeled data.
A question about profit fade answered from certified measures, naming two jobs and citing the measure, the WIP schedule and the forecast it used.
Sample company with modeled data.
A project summary showing contract value, earned and billed to date, buyout, margin at estimate against forecast, fee at risk and the cost codes driving fade.
Sample company with modeled data.
The cash forecast for a CFO: $36.0M at 13 weeks within about $1.3M, high confidence; $39.4M at 26 weeks within $2.6M; $43.5M at 52 weeks within $5.4M, lower confidence. Error at 52 weeks has fallen from 19% to 12% over four closes as the model learns.
Sample company with modeled data.
The company pulse for a CEO: a volatility index of 56, in the normal band, with schedule slip moving most, and a score for each region, one of them elevated.
Sample company with modeled data.
Forecasting
Know where the books will land before accounting makes a single posting.
No more lagging indicators, only leading ones. dideh forecasts revenue, margin and cash from the work itself: every job’s schedule, cost to complete, billing terms, retainage and pay-when-paid. It rolls them up into the three statements, tied out to your books, so on the last day of the month you already know what the close will say.
Day 0
dideh
Where the month lands: revenue, margin and cash, from the work itself.
Day 5
Invoices posted
Subcontractor and supplier bills entered.
Day 10
Job cost reconciled
Cost to date agreed to commitments.
Day 15
WIP schedule
Over and under billings worked out.
Day 20
Statements issued
Leadership finally sees the month.
Financials · Sample GC
Forecast
Monthly revenue. Actual for Oct ’25 to Sep ’26, rising from $30.6M to $37.9M. Forecast for Oct ’26 to Sep ’27, dipping to $32.6M in January and ending at $35.8M.
| Measure | Q4 ’26 | Q1 ’27 | Q2 ’27 | Q3 ’27 |
|---|---|---|---|---|
| Forecast | $104.7M | $100.9M | $106.6M | $107.8M |
| Plan | $108.9M | $102.6M | $107.4M | $111.1M |
| Against plan | −$4.2M | −$1.7M | −$0.8M | −$3.3M |
| Gross margin | 15.4% | 15.7% | 15.9% | 16.1% |
The year forecasts $10.0M under plan. Most of the gap sits in Q4 ’26, where two jobs are running late, and in Q3 ’27, where less of the work is signed.
Sample company with modeled data. On the cash forecast, change how owners pay and watch the low point move.
Each job forecasts from its schedule, cost to complete and buyout, so the numbers move the day a slip or a change order happens.
Job forecasts roll up into a forecast income statement, balance sheet and cash flow that tie out like your actuals.
Actual, forecast and plan side by side, by month and quarter, for the company, each region and each job.
See what slower-paying owners would do to your low point, your minimum and the revolver before it happens.
Analytics
The analysis a data science team would build, run on your own jobs and kept current: where each margin is heading as the work progresses, how your estimates held up on finished work, and how the regions compare. Each chart comes with its takeaway written out.
Open jobs · All regions
Each square is an open job: how far along it is on cost, the margin it is heading for, and its size. The shaded corner holds jobs past halfway and below the average bid, with the least runway to recover.
A scatter of 48 open jobs by percent complete and estimated margin. 18 are past halfway and below the 17.0% average bid.
6 jobs past halfway have faded more than 1.5 points. Alder Stone Compound (13.9% against a 16.3% bid at 55% complete), Lakeshore Penthouse (14.9% against a 17.0% bid at 62% complete) and 4 more carry $3.1M of fade with less than half the cost left to recover it.
Open jobs · All regions
Open jobs by the margin they are heading for, in one-point bands. Bands under the average bid are picked out.
32 of 48 jobs, 66% of contract value, are heading below the 17.0% average bid. The median job sits at 16.5%.
Completed jobs · 2024 to 2026
Each square is a finished job: the margin it was bid at against the margin it closed at. Jobs under the diagonal finished below bid.
Final minus bid, points
86 completed jobs plotted by bid margin against final margin. Final minus bid: −4.0 points, 2; −3.5 points, 2; −3.0 points, 4; −2.5 points, 9; −2.0 points, 10; −1.5 points, 13; −1.0 points, 10; −0.5 points, 13; at or above bid, 23.
63 of 86 finished below bid. Value-weighted, completed work closed at 16.6% against a 17.6% bid (−1.0 pts, −$10.8M). GMP −1.3 pts · Lump sum −1.4 pts · Cost-plus 0.0 pts. The open book is tracking −0.6 pts so far.
Regions · Trailing twelve months to Sep 30
Results, runway and win rate for each region, with its fiscal 2027 forecast. The regions add up to the company, and each total ties out.
| Region | Revenue | Monthly trend | Growth | Gross margin |
|---|---|---|---|---|
| Pacific | $104.6M | +13.1% | 16.5% −0.6 | |
| Mountain | $68.2M | +6.4% | 17.4% +0.1 | |
| Southeast | $86.5M | −5.2% | 16.1% +0.2 | |
| Central | $77.9M | +30.4% | 15.5% +0.3 | |
| Northeast | $64.6M | +20.3% | 14.9% −1.4 | |
| Company | $401.8M | +11.2% | 16.1% −0.3 |
Gross margin change is year over year, in points.
Fade on open jobs, by region
Read across. Pacific is the largest region at $104.6M (+13.1%). Central grew fastest (+30.4%) on the shortest runway, 13.0 months of backlog, and Southeast is the only region shrinking (−5.2%). Mountain earns the best margin (17.4%) and Northeast the thinnest (14.9%, down 1.4 points), with $1.8M of fade on open jobs.
Sample company with modeled data.
Autopilot
Each specialist has a written mandate, a person accountable for it and access to only what that mandate needs. It watches the measures it owns, proposes the next move with evidence and a draft, and acts only as far as you allow.
Keeps the 13-week cash low point above your minimum, and proposes a revolver draw before a shortfall.
Keeps owner billings and retainage on the clock: drafts reminders and escalations, without chasing payments owners have already scheduled.
Finds work in place that isn’t billed and change orders aging past 30 days, and drafts the fix.
Catches fade while there is still runway: reforecasts, closeout reviews and recovery from vendors.
Proposes pay runs for compliant vendors once the owner has paid, and flags any vendor without current insurance or lien waivers.
Recommends the best risk-adjusted vendor for every package before its buyout date, and keeps late deliveries off the critical path.
Protects substantial completion: recovers the days your trades lost where it pays, and documents the days the owner cost you.
Keeps crews where the backlog needs them and every project manager inside a span they can run.
Points preconstruction effort at the pursuits you can win, and recommends the bid margin where expected profit peaks.
Gives every high risk an owner and a next step, and finds the concentration no risk register shows.
Sequences the specialists, flags conflicts between them and briefs the CEO on what was done and what needs a decision.
The next specialists will be built with our founding partners, around how they run.
Become a founding partnerEvery number a specialist cites comes from a certified measure, and every proposal, decision and action is logged with its evidence.
One setting per specialist, and per action type if you want. Every specialist starts in simulation: you see what it would have done, and what that was worth, before it does anything.
Proposes each move with its evidence and a draft. A person decides.
For example, the Billing specialist
Billing drafts the revised pay application for Harbor View and routes it to the CFO, because $412K is above what the VP of Construction can approve.
Sense
Watches WIP, cash, buyout, schedule, labor and risk through certified measures.
Reason
Connects the exception to the job, the owner, the vendor and prior decisions.
Route
Checks the amount, the action type, the specialist’s limits and who has the authority to approve.
Decide
The accountable leader approves it, sends it back or asks for more. Every step is logged.
Billing specialist · Harbor View Residence
Revise the September pay app to bill $412K of earned work.
Above the $250K the VP of Construction can approve, so it is routed to the CFO.
Sample company with modeled data.
Every seat
People and specialists see what their role allows and approve what their authority covers. Pick a seat and watch one job change.
Job 24-118
Harbor View Residence
Seen asCEO
CEO: every value on this job is visible.
Example policies on the sample company. Yours are set by role, region and job.
Trust
dideh governs who can see a number, which measures are certified, what a specialist may do and when a person must decide. And it can run in your own Azure tenant.
People and specialists
Today · My Decisions · Ask dideh · Autopilot · your own AI tools
Governance
Seats · masking · approval authority · limits · audit
Context layer
Construction concepts · certified measures · knowledge graph · documents
Your systems
ERP · project management · schedule · field · CRM · payroll · documents · bank
Runs in your Azure tenant
On request, dideh is deployed inside your own Microsoft Azure environment.
Single sign-on
People sign in through your identity provider, such as Microsoft Entra ID or Okta.
SOC 2 status
Our SOC 2 status and documentation are available to review under NDA.
Your AI tools, same rules
Your own AI agents can query the governed model over the Model Context Protocol, under the same seat policies.
Least-privilege specialists
Each specialist runs as its own service account and reads only what its mandate needs.
A full audit trail
Every query, answer, proposal and decision is logged: who asked, what was read and who decided.
Read-only, with lineage
Nothing is written back to your systems, and every number traces to the field it came from.
Your data stays yours
An NDA before anything is shared, a dedicated, isolated database, export at any time, and AI providers whose terms exclude training on your data.
Services
The platform runs on clean, connected data and a finance function ready to act on it. When either isn’t there yet, our CFO-led advisory assesses the business and its systems, and our data team cleans and connects the data, so dideh goes live on numbers leadership trusts. Start with any layer.
An independent, CFO-level read on the numbers and what to do about them, with a CTO-level read on the systems and data behind them. We assess and advise; your team keeps running day-to-day finance.
| Area | What we assess | What you receive |
|---|---|---|
| CFOFee and pricing | True margin by project type, region and PM, the labor billing-rate spread, and win-loss patterns | Fee and billing-rate recommendations, refreshed quarterly |
| CFOFinance and accounting team | The department and each team member: roles, capacity, skills and accountability | A private assessment, org recommendations and hiring guidance |
| CFOClose and reporting | The monthly close and the management and ownership reporting your team prepares | A tighter close, and a monthly review of results with leadership |
| CFOCash and capital | The cash forecast, liquidity, and bank, surety and auditor relationships | A quarterly cash and capital assessment |
| CFOProcess and controls | Finance processes and controls, including bottlenecks and manual handoffs | A controls assessment with recommended policies |
| CTOTech stack | Every system that touches the numbers: what each one holds, how they connect and who can see what | A systems map and the order to close the gaps in |
| CTOData integrity | Job numbers, cost codes and vendor records across systems: where they disagree, and why | A data-readiness read and a cleanup plan, sized from sample rows under NDA |
The first 90 days
Days 1–30
Meet the finance team; review the close, reporting, cash, and the systems behind them.
Days 31–60
Team, process and data assessments for leadership, in confidence, with quick wins.
Days 61–90
First fee and pricing findings, and a 12-month roadmap for finance and data.
Advisory only: we don’t manage staff, sign filings or hold bank signing authority. Month to month, with hours reported monthly.
Ask about advisoryReadiness to go live
Illustrative
0 of 6 ready
Fee targets set by project type and region, not ready yet
Closed by CFO advisory
A close on a set calendar, with reporting leadership trusts, not ready yet
Closed by CFO advisory
Every system mapped: what it holds and who can see it, not ready yet
Closed by CTO advisory
One job key from estimate to closeout, not ready yet
Closed by Data foundation
Schedule activities mapped to cost codes, not ready yet
Closed by Data foundation
Job cost tied out to the ledger, not ready yet
Closed by Data foundation
Already have clean tables or Power BI? We connect to them rather than rebuild.
How we start
dideh sits on top of the systems you already run, with nothing to migrate. If your data is clean, connecting is close to plug and play, and leadership sees each view as soon as it’s ready.
Data not ready yet?
CFO-led advisory assesses the finance function and the systems behind it, and our data foundation team cleans, maps and ties out the data, so the platform goes live on numbers leadership trusts.
See the servicesAn NDA, then a sample of rows from each system. Within a week we tell you whether your data is ready to plug in, or exactly what needs cleaning up first, which our data foundation team can take on.
Each system connected read-only, one job key across all of them, and every view tied out to your accounting: fee margin, WIP, cash forecast, AR and AP.
Schedule activities mapped to cost codes, for progress against cost burn, forecast at completion and fee at risk.
Today, Ask dideh and the dashboards live for leadership, with a fee-margin readout by job, region and project manager.
Specialists come on one at a time and start in simulation, so you see what each would have done and what it was worth. Then you choose how far it goes: watch, recommend or act inside limits.
Founding partners
We’re taking on a small number of general contractors as founding partners. You get a first phase scoped with us, and a direct hand in what dideh learns to do next.
For general contractors that run an ERP and a project management system, and want their leadership working from one version of the numbers.
FAQ
If yours isn’t here, bring it to the walkthrough.
No. dideh connects to the systems you already run, read-only, and they stay the source of record. It adds one governed model of the company on top, so the numbers agree everywhere.
From certified measures, each defined once with an owner and tied out to your accounting system. A number in Today, in Ask dideh or in a specialist’s proposal is the same number.
Revenue, gross margin and operating income by month and quarter; cash week by week for 13 and 26 weeks, with the low point against your minimum; every open job’s margin at completion; and how much of the forecast is already under contract. Each forecast is built from the work itself, from schedules, cost to complete, billing terms, retainage and pay-when-paid, and rolls up into a forecast income statement, balance sheet and cash flow that tie out to your books.
Every seat is governed by role, with row filters, masking and approval authority, and each AI specialist is a seat too. For example, a superintendent can be kept from seeing margin, and the Collections specialist from reading salaries.
Only as far as you allow. Each specialist can watch, recommend, or act on internal moves inside dideh within limits you set, such as ordering a reforecast or opening a billing review; nothing is written back to your systems. Anything that reaches an owner, a vendor or a bank, anything over its dollar limit and anything it isn’t confident enough about goes to a person with the authority to approve it. Every step is logged, and specialists start in simulation.
Yes. On request, dideh is deployed inside your own Microsoft Azure environment.
Yes. Your own AI agents can query the same governed model over the Model Context Protocol, under the same seat policies as your people, so they work from certified numbers.
No. dideh’s answers are built from your data, and the AI providers it uses have commercial terms that exclude training on it.
Bring your security team to the walkthrough. People sign in through your identity provider, such as Microsoft Entra ID or Okta, and we share our SOC 2 status and documentation under NDA, along with how dideh is deployed and how access is governed.
If your data is clean, it’s close to plug and play: dideh connects to the systems you already run, with nothing to migrate. In the first week we check a sample from each system and tell you what, if anything, needs cleaning up; if it does, our data foundation team cleans, maps and ties it out to your ledger. Leadership sees each view as soon as it’s ready.
No. Each layer stands on its own: some teams start with CFO advisory or the data foundation alone, others plug straight into the platform. Each one makes the next one stronger.
No. Advisory is an advisory role: your team owns day-to-day finance, staff and relationships. We assess, advise and review results with leadership, and we don’t sign filings or hold bank signing authority.
One of a small number of general contractors working closely with us: a first phase scoped together, specialists tuned to how you run, and a say in what we build next.
dideh running on a fully modeled general contractor: WIP, forecasts and pay-when-paid cash, the analytics behind them, questions answered for each seat, and the specialist team in simulation. It is all sample data; nothing comes from any client.
Walkthrough
We’ll walk you through a fully modeled general contractor, with WIP, pay-when-paid cash, governed questions and the specialist team in simulation. Then we’ll talk through your systems and what a first phase would cover.
Prefer email? Write to hello@dideh.ai.
