Key takeaways
- Talk to your office team before any AI tool goes live, and be honest about what you have and have not decided.
- 87.6% of construction firms using AI said it had not changed their headcount over six months[1].
- BLS still projects about 105,100 receptionist openings a year despite declining employment[2]; the work changes more than it disappears.
- Give staff a real role in the pilot: testing, reviewing calls and owning the rules.
What your team is probably thinking
When an owner says the company is trying AI on the phones, the people who answer those phones hear one question: is my job safe? They may not ask it out loud. Instead you may see quiet resistance, a rush of examples of the system getting things wrong, or a good CSR updating their resume. None of that is irrational. It is what people do when a change affecting their livelihood is announced without explanation.
The usual worries in a plumbing office are specific:
- I'll be replaced, or my hours will be cut.
- I'll be blamed when the AI gets something wrong.
- I'll spend my day cleaning up its mistakes.
- Customers will be angry and I'll take the calls.
- My experience with our customers won't count any more.
- Nobody asked me, even though I know how these calls really go.
Every one of these can be answered, but only if you have thought about it first.
What the data actually shows
So far, AI adoption in construction has not been a story of mass layoffs. Among construction businesses using AI, 87.6% said it had not changed their headcount over six months, 4.2% said it increased and 3.1% said it decreased[1]. In a U.S. Chamber of Commerce survey, 82% of small businesses using AI said they had increased their workforce over the past year[3], though growing firms may simply be more likely to adopt new tools.
Be honest about the longer trend, too. The Bureau of Labor Statistics projects employment of customer service representatives to decline 5% from 2025 to 2035 as their tasks continue to be automated[4], and receptionist employment to decline 2%[2]. Yet it still projects about 105,100 receptionist openings each year, mostly to replace people who leave[2]. The jobs are changing shape rather than vanishing, and the people who understand both the customers and the tools will be the ones in demand.
Decide before you talk
Before the conversation, be clear with yourself about what you are trying to achieve and what you will promise. Staff can handle uncertainty; they cannot handle discovering later that you knew more than you said.
- What problem are you solving? Missed after-hours calls, lunchtime overflow and slow callbacks are easy to explain and rarely threaten anyone's role.
- What will you commit to? If you are not planning to cut roles or hours, say so plainly. If you have not decided, say that, and say when you will.
- What will change for each person? Write it down per role before you meet.
- Who will own the system day to day? Ideally someone on the office team.
Do not promise that no job will ever change. Promise what you can keep: how decisions will be made, and that the team will hear first.
The conversation, step by step
Meet in person, as a team
Tell the office team together, before anyone hears from a vendor or a technician. Follow up one to one with each person within a week.
Start with the problem
Show the numbers: calls missed at night, voicemails that never turned into jobs, the hours spent on callbacks. People accept a solution more readily when they agree on the problem.
Explain the scope
Say exactly what the AI will and will not do at first, for example nights and weekends only, with bookings held as requests for the office to confirm.
Answer the job question directly
State your commitment on roles and hours, or when you will decide. Do not leave it to be inferred.
Give them a role
Ask for volunteers to test the system, review calls and own the rules. Their knowledge of real calls is the best quality control you have.
Set a review date
Agree when you will all look at the results together, and keep that date.
Lines that help, and lines that hurt
- Helps: We're losing jobs at night because nobody answers. This is about those calls.
- Helps: You know our customers better than any software. I need you to tell me when it gets things wrong.
- Helps: Nobody's hours are being cut because of this. If that ever changes, you'll hear it from me first.
- Helps: I want you to spend less time returning voicemails and more time on the calls that need a person.
- Hurts: It's basically going to do your job, but cheaper.
- Hurts: Don't worry about it.
- Hurts: The vendor says it's better than a human.
How the work changes
| Task | Before | After |
|---|---|---|
| Night and weekend calls | Voicemail; callbacks next morning | AI answers; office reads summaries |
| Routine bookings | CSR books every call | AI books simple jobs by rules; CSR handles exceptions |
| Reschedules | CSR handles by phone | AI handles simple moves; CSR handles conflicts |
| Complaints | CSR or manager | Still a person, now with the recording |
| Rules and answers | In people's heads | Written knowledge base owned by the office |
| Quality | Owner's impression | Weekly review of call samples by the team |
The work that stays is often the most valuable: calming an upset customer, sorting out a billing question, coordinating a difficult job with a property manager, spotting that a caller is elderly and confused. Make that explicit, and consider whether new responsibilities, such as owning the knowledge base or reviewing quality, deserve recognition in pay or title.
Training for the first month
People accept a new tool faster when they understand how it works and how to correct it. Plan a short, practical training rather than a vendor slideshow:
- A one-hour session where staff make test calls themselves, including the awkward ones they know customers make.
- A walk-through of where summaries, recordings and transcripts appear, and how to search them.
- How to report an error: one shared place, one line per issue, no blame.
- How to change an approved answer or ask the owner of the knowledge base to change it.
- What to say when a customer asks whether they spoke to a robot, so everyone gives the same honest answer.
Repeat a short version for anyone who joins later. A new CSR who arrives after go-live should learn the AI rules alongside the price book and the service area, not pick them up by accident.
Technicians need to hear it too
The office is not the only group affected. On-call technicians will receive urgent calls routed by the system, often at night, and they will be the first to notice if callers are described badly or routed to the wrong person. Brief them separately: what will reach their phones, in what form, and what to do if something looks wrong. Ask them to report problems the same night. Techs who feel the system was imposed on them tend to ignore its summaries and call customers cold; techs who were asked for their input tend to read them and arrive better prepared.
Example: one office, three roles
Example: a shop with two CSRs and a dispatcher introduces AI for after-hours and lunchtime overflow. The senior CSR becomes owner of the knowledge base and spends Monday mornings updating it from the weekend's call records. The second CSR reviews five calls a day and flags errors, and now spends the time she used to spend on voicemails following up open estimates. The dispatcher starts each day from summaries rather than voicemails and reports that urgent jobs arrive with clearer addresses. Six months later nobody has left, and the senior CSR has a new title and a raise.
Signs it is going badly
- Staff stop reporting errors, or only report them to each other.
- The system is blamed for problems it did not cause, or never blamed for ones it did.
- Review time keeps growing instead of shrinking.
- Good people start asking about hours or leaving.
- Customers mention staff saying negative things about the new phones.
If you see these, pause any widening of scope and talk again. The tool can wait; a good CSR is harder to replace than any software.
Frequently asked questions
Should I tell staff before trying AI?
Yes, before anything goes live and before they hear about it from someone else. Explain the problem, the scope and your commitment on roles and hours.
Will AI replace my CSRs?
In most plumbing shops it takes repetitive intake and after-hours calls, while people handle complaints, exceptions and complex jobs. Survey data shows most construction firms using AI have not changed headcount.
How do I get staff to support it?
Give them a real role: testing, reviewing calls and owning the knowledge base. Their knowledge of real calls makes the system better, and ownership reduces fear.
What if a staff member refuses to work with it?
Listen first; resistance often points to a real problem. Then explain what is expected, offer training and give it time before treating it as a performance issue.
References
U.S. Census Bureau (2026). Business Trends and Outlook Survey (BTOS), response estimates by sector (opens in a new tab) โ
U.S. Bureau of Labor Statistics (2025). Occupational Outlook Handbook: Receptionists (opens in a new tab) โ
U.S. Chamber of Commerce Technology Engagement Center (2025). Empowering Small Business: The Impact of Technology on U.S. Small Business (2025) (opens in a new tab) โ
U.S. Bureau of Labor Statistics (2025). Occupational Outlook Handbook: Customer Service Representatives (opens in a new tab) โ