Superintelligence Prep for Hardwood Refinishers: Quotes and Scheduling

An AI agent for a hardwood floor refinisher can follow up on open estimates, confirm crew start dates and chase signed approvals, but only when the quote record states the floor type, finish system, measured area, site conditions and the crew calendar rules that govern each job. Without those fields, the agent has to guess, and a guess about sanding dates or final cost is the fastest way to lose a sale that was almost closed. This article covers what to capture before any agent touches your pipeline.

Key takeaways

• An agent can only follow up on an estimate it can read in full, so the quote record has to be fixed before the agent is introduced.
• Scheduling rules such as sanding passes, screening, coats and cure time must be written as constraints, not held only in a lead installer's head.
• Price changes, scope exclusions, deposit terms and warranty language stay with a person until the approval boundary is written down.
• Every follow-up should leave a logged outcome so the team can see which estimates stalled and why.
• The first useful agent task is usually narrow, such as chasing estimates older than a defined window, not running the whole sales process.

Why follow-up fails before the agent writes a single message

Most refinishing estimates are made on site. The estimator takes notes on a phone, sketches the rooms, and records a verbal description of the finish the client wants. The office later types a figure into a quote template. By the time a follow-up is due, the record may say "oak, 1,200 square feet, two coats" and nothing about the stair nosings, the pet staining in the hallway, or the client's request to leave the living room furniture in place. An agent reading that record can only send a generic reminder. Generic reminders are easy to send and rarely close jobs.

Refinishing also has its own vocabulary: the grit sequence, screening versus full sanding, stain or no stain, water-based or oil-based topcoat, the number of coats, and how long the floor needs before furniture returns. Each of these changes the schedule and the risk. A record that omits them forces either a person to re-check every follow-up or the agent to improvise. Improvising about floors is where a business loses credibility.

Which fields belong in a refinishing quote record

The record does not need to be long, but every field should be filled in or explicitly marked as unknown before the estimate is sent. At minimum, capture:

• Species, board width if known, and whether the floor is solid or engineered.
• Condition notes: gaps, cupping, pet staining, previous finish type, and any repairs that must happen before sanding.
• Measured area and the basis of measurement, including closets, stairs, landings and transition strips.
• The finish system: sanding passes, stain choice, topcoat type, sheen and coat count.
• Site access: which rooms stay occupied, pets, dust containment needs, and who moves furniture and when.
• Cure and return-to-use times as written in the job notes, with the rule that produced them.
• The estimate's validity date and the role allowed to change it.
• Decision status, last contact date, and the exact question the client has not yet answered.

The last field matters most for follow-up. A message that says "Are you still deciding between matte and satin topcoat?" is useful. A message that says "Just checking in" is not. If you sell painting alongside refinishing, the house painters guide on estimates and color records covers a similar record structure for a neighboring trade.

How do crew calendar rules turn into scheduling constraints?

A refinishing crew calendar is not a list of appointments. It is a sequence with waits built in. Sanding has to finish before stain, stain has to dry before topcoat, and topcoat needs cure time before foot traffic or furniture returns. Some rooms become unusable for a set period. Humidity and weather can move the plan, and dust from one room can reach another if containment is weak.

The practical move is to write these dependencies as short rules attached to job types, so the agent can read them before it proposes a date. A rule might say that a full sand requires the rooms to be vacated, that a screen-and-recoat job can run with light furniture in place, or that a stair-only job cannot be confirmed without a site visit. The agent does not need to understand flooring chemistry. It needs to know which conditions must be true before it can say a date is available.

The calendar is only as reliable as the constraints behind it. A date that looks open may be blocked by a cure window nobody wrote down.

Illustrative scenario: A crew has an open slot on a Monday in a townhouse. The estimate covers screening on the upper floor only, but the client has since asked for full sanding on the stairs. The record has no rule for stair work, and the calendar shows no buffer for it. The agent should not confirm the slot. It should hold the follow-up, flag the scope change to the office, and ask a person to update the estimate before any date is offered.

A step-by-step plan to build the record before the agent

You can do this work with the staff you already have. The sequence below keeps the agent's first role small and reviewable.

• Choose one source of truth for estimates, usually the job management or CRM record, and stop keeping parallel notes in messages and personal files.
• Add the fields listed above to the estimate template, marking which ones are required before an estimate can be sent.
• Write scheduling rules as short statements for each job type, including the cure window and who may approve an early return to use.
• Attach those rules to job types so the agent reads the right constraints for a screen, a full sand, or a stair-only job.
• Decide which estimates the agent may follow up on alone, and which require review, such as any change in scope after the site visit.
• Log every follow-up with the date, the rule that applied, the message category, the client's reply and the next step.
• Review the log on a regular schedule and tighten the rules wherever the same question keeps coming back.

Seasonal demand creates a second pattern worth planning for. Outdoor trades handle rebooking differently, and the lawn care quotes and routes article shows how recurring work can be tracked. Refinishing is usually one-time work, so the follow-up logic should focus on decisions and calendar holds rather than recurring visits.

What the agent should never change on its own

The permission list should be short and written down. An agent working on estimates should not change or promise any of the following without a named person's approval: the price or any line item, scope exclusions, deposit or payment terms, dust control or pet safety promises, warranty language, and any crew start date that conflicts with an existing booking. It also should not accept a new scope from a client message and write it into the record as if it were approved.

Enforce this with access rules rather than instructions alone. The OWASP authorization guidance describes checking each action against the caller and the object being changed, so that a follow-up agent can read the estimate and propose a message but cannot edit the price field. Client emails and form submissions also need to be treated as data, not as instructions. The OWASP prompt injection guidance explains why text pasted into a customer message can try to redirect an agent, and why the agent's permissions should not depend on what a message says.

How do you know the follow-up is working?

Sending messages is not the same as observing outcomes. The OpenTelemetry observability concepts describe telemetry as the signals that let you understand what a system is doing from its outputs. For an estimate agent, those signals are simple: which estimate was touched, which rule was applied, what was sent, whether the client replied, whether a person stepped in, and what happened to the job. A weekly review of that log shows whether stalled estimates are stalled because of price, timing, or a scope question the record never answered. That last category usually points to a missing field in the template.

Agent design guidance from Anthropic favors simple, well-bounded workflows over open-ended autonomy when the task has clear rules. Estimate follow-up fits that description well, which is why it is a common first role.

How Be Human's managed approach handles this

Be Human starts with one useful role, such as an estimate follow-up and crew confirmation agent for a refinishing business. The first engagement defines the memory the agent needs: the estimate record fields, the rules attached to each job type, and the crew calendar constraints. It sets approval boundaries for price, scope, deposits and booking changes, and it defines an observable success criterion, such as every follow-up being logged with its rule and outcome. After that, the managed service covers monitoring, recurring evaluations of the agent's messages against the written rules, permission reviews as staff and job mix change, integration maintenance with the estimate and calendar tools, and controlled improvements. The service is delivered remotely, in English or Portuguese.

If you want to find the first task worth handing to an agent, take the free Be Human AI diagnostic.

Frequently asked questions

Can an AI agent follow up on hardwood refinishing estimates without a person reviewing each message? It can, for a narrow set of estimates where the record is complete and the message only asks a question the record already covers. Scope changes, price adjustments and any message that promises timing or results should still go to a person. Starting with that boundary keeps the first version easy to review.

What should we do if our estimates were written from a phone photo and a verbal description? Treat that estimate as incomplete and mark the missing fields before any follow-up is sent. A short site revisit or a structured call can fill most gaps in one step. Once the template enforces the fields, the agent can work from a consistent record.

Should the agent confirm crew start dates directly with clients? Only when the calendar rules for that job type are written, the slot is free after cure and buffer time, and no scope change is pending. In every other case the agent should hold the date and route it to the scheduler. Confirming a date that a rule would have blocked is the failure to avoid.

How long does it take to get ready for this kind of agent? The timeline depends on how much of the record already exists and how many job types you run. Most businesses start by writing rules for one or two common job types and expanding once the log shows what the agent handles well. Be Human's first engagement is built around that scoped start rather than a full rebuild.

Be Human editorial guidance. Examples are illustrative unless explicitly attributed. Sources provide background; implementation choices depend on your workflow.