DeskForeman answers texts in seconds, remembers project details, follows up with stale leads, prepares estimate inputs, routes builder questions, and keeps approved next steps moving. You stay in control.
Maya starts signup by text and sends the Stripe setup link in the thread. Live customer traffic still waits for approval-gated onboarding.
Other tools are reactive — they respond when a message comes in. DeskForeman follows up, nudges, surfaces approvals, and keeps your pipeline organized without adding another dashboard habit.
4 active nudges when a lead goes quiet, then monthly check-ins. No lead falls through the cracks.
Parked leads get warm, contextual check-ins at day 14, 30, 60, and 90. No pressure, just presence.
Detects customer frustration and texts you immediately with context.
Alerts about pending approvals, stale leads, and action items.
Biweekly SMS with conversion rates, stale lead alerts, and coaching.
Prompts for measurements and scope notes to draft the proposal.
of homeowners won't hire a contractor
who missed their first call
of homeowners say communication — not price —
is their biggest complaint with contractors
DeskForeman is your AI Front Office. It talks to your customers, follows up proactively, qualifies leads, and keeps you in control.
What your customer sees
What you see
From first text to builder-approved next step. You stay in control.
Organizes proposal and contract handoff details from site visit data. Customer-facing document sends stay gated until the builder has approved the right path.
Live status on stale leads, approvals, builder questions, and queued handoffs. DeskForeman keeps the next step visible.
Formula-anchored estimate drafts using per-sqft industry rates, regional labor assumptions, and material adjustments. Customer-visible price ranges stay behind builder approval.
Every lead scored 0-100 across six dimensions: location, project fit, budget, timeline, specificity, and engagement. Configurable thresholds route leads to qualify, nurture, or decline — automatically.
Collects visit availability and routes a clean schedule request for builder approval.
Replies in seconds, 24/7. No lead left waiting while you're on a job site.
Every customer-facing response is reviewed for compliance, pricing accuracy, and tone before the customer sees it. DeskForeman doesn't go rogue.
Most AI tools are a prompt and a prayer. DeskForeman's lead scoring and estimation are formula-driven systems that get smarter with every project.
Example: Sarah's composite deck project
Biased toward qualification — a false positive costs a conversation, a false negative costs a $45,000 job. You set the thresholds. DeskForeman scores every lead the same way, every time.
How every estimate is calculated
Per-sqft base rates for every project type. Material multipliers for standard, premium, and luxury finishes. Site difficulty adjustments for slope, access, and elevation. Then your margin on top.
Answering fast is table stakes. Staying engaged is where contractors lose deals — and where DeskForeman takes over.
homeowners say getting timely updates from their contractor
was the hardest part of the project
per week builders spend on admin that has nothing
to do with building — texts, follow-ups, scheduling
Most AI tools give you a settings panel. DeskForeman earns trust the way a real employee does — through your approval patterns. You stay in control, and it gradually takes more off your plate at your pace.
Prices, schedule confirmations, proposals, contracts, warranties, deposits, permits, and material-policy claims go through the builder first.
Routine tasks handled. Unusual items flagged for review.
DeskForeman runs your pipeline. You get alerts only on exceptions.
"You've approved 8 out of 9 estimates I flagged. Want me to start sharing ballpark ranges on my own? I'd still flag anything unusual."
This isn't a toggle. It's earned delegation, tracked across every interaction. No other AI tool works this way.
From first text to builder-approved next step, in one AI Front Office.
Most builders have. Here's how DeskForeman is different.
Ruby, Smith.ai, etc.
They answer the phone. Then what? DeskForeman runs the front-office workflow from first text to builder-approved next step. An answering service captures a name. DeskForeman keeps qualified opportunities moving.
Jobber, JobNimbus, etc.
Dashboards only work if you open them. 30-40% of contractors stop logging in within months. DeskForeman runs on text — zero behavior change required.
Drift, Intercom, etc.
Chatbots follow scripts. DeskForeman uses deterministic workflow rules, prepares estimate inputs from trade data, and keeps risky claims behind approval. There is no comparison.
Each connection makes DeskForeman more valuable without increasing the price.
DeskForeman collects availability and routes visit times for builder approval before anything is confirmed.
Live conversations, approvals, pipeline state, and operator watch data stay visible during the pilot.
New leads auto-create customers and jobs in your CRM. No double entry.
Approved jobs can later flow into accounting as estimates or invoices. Books stay current.
The first production tenants are hands-on and approval-gated. We watch the first live conversations, keep builder decisions strict, and tune Maya against real contractor workflows before broad signup.
The average residential project is $5,000-$80,000. One recovered lead can pay for a lot of front-office help.
Signup starts over SMS. Maya collects the basics, then texts the Stripe setup link. Pilot onboarding stays hands-on.
| DeskForeman | AI Receptionists | Contractor CRMs | |
|---|---|---|---|
| Pipeline coverage | First text to approved handoff | Answering only | Manual, if you log in |
| Lead scoring | 6-dimension weighted model | None | Manual tags |
| Estimate and proposal prep | Formula-anchored + builder approval | None | Manual templates |
| High-risk claims | Builder-approved only | Scripted | Manual |
| Builder manages via | Text messages | Dashboard | Dashboard + app |
| Setup | Hands-on pilot onboarding | Script config | Days-weeks |
SMS-first signup, Stripe setup link in the thread, operator-watched onboarding before live customer traffic.
The Big Picture
Every customer message flows through the same intelligent pipeline, no matter the channel.
Texts, calls, or emails
SMS, voice, email, web
deterministic checks in real time
Alerts, approvals, control
DeskForeman's AI Tasks
Specialized AI models matched to each task. Reasoning-heavy work gets frontier models. Speed-critical checks use fast models at 10-15x lower cost.
Talks to customers naturally via SMS and phone. Gathers project details, answers questions, and extracts lead data in a single conversation turn.
Scores every lead 0-100 across six weighted dimensions: location, project fit, budget, timeline, specificity, and engagement. Configurable thresholds route leads as Qualify, Park, or Disqualify.
Formula-anchored estimate drafts using per-sqft base rates, regional labor assumptions, material and site adjustments. Customer-visible ranges include confidence level and assumptions after builder approval.
Prepares proposal-ready notes with scope narratives, inclusions, warranty terms, and payment schedule inputs for builder approval. Contract handoffs stay gated until the builder has an approved document path.
Prepares site-visit handoffs: availability, customer context, builder approval, and a prep checklist before anything is confirmed.
Every response is reviewed for policy compliance before sending. A separate detector monitors for customer frustration and alerts you via SMS if things escalate.
The Automated Pipeline
Each step follows deterministic prerequisites. High-risk claims stay behind builder approval.
Customer texts. DeskForeman responds in seconds.
Lead scored. Hot leads flagged to you.
Ballpark price from real market rates.
Site visit request ready for approval.
Document details organized.
Builder controls the send.
Next step is clear.
Engineering
Reasoning-heavy tasks run on stronger models. Speed-critical checks and deterministic gates stay lightweight, auditable, and repeatable.
The customer assistant returns a conversational response and structured lead data in one API call, cutting latency and cost in half versus a two-call pattern.
Background automation only fires when new information is shared. Chitchat doesn't trigger the qualifier or estimator, preventing wasted compute and unnecessary builder notifications.
Circuit breakers on every external dependency. Exponential backoff retries. Graceful degradation: if the AI API goes down, customers get a friendly hold message, not silence.
Stack
Right model per task
Node 22, TypeScript
SMS, voice, dedicated numbers
API, worker, site
Multi-tenant, read replicas
Outbound dispatch
Operator control
Pilot checks
Multi-tenant architecture · Deterministic workflow gates · Worker heartbeat · TCPA/SMS compliance · Watched pilot operations