The category guide
The category guide: what an AI back office is — and isn't
The office is the bottleneck nobody budgets for
Ask a service-business owner what stops growth and few say "not enough work." They say: the phone, the invoices, the follow-ups, the scheduling, the paperwork — the office. The owner becomes the office at night. The traditional fixes are hiring (slow, expensive, fragile — one resignation from chaos) or software (necessary, but software is a tool cabinet, not a worker). The AI back office is the attempt at a third fix: a team that does the work, at software economics.
What a real AI back office covers
- Demand in: answering, booking, estimate follow-ups, review requests, reactivating quiet customers.
- Work organised: scheduling and dispatch support, job records, crew communication, reminders.
- Money: invoicing on completion, chasing late payments, knowing per-job cost and margin.
- People & records: customer history in one place, staff records, documents findable.
- The owner's view: a morning picture of the business — what's booked, what's owed, what needs a decision — and a longer view: what would make this business worth more.
Software vs service vs teammates — the honest taxonomy
Platforms (ServiceTitan, Housecall Pro, Jobber) are where the records live. You need one. Their AI features increasingly help the human operating them. They do not, today, operate themselves.
Managed AI services put a provider's people and AI between you and the work. Fast to start, genuinely hands-off — and the memory of how your office runs accumulates in the provider's organisation, priced monthly like rent.
AI teammate suites (Deck 72's shape) put the team inside your business: named roles, one shared memory that belongs to your shop, your approval law, your data. The newest shape, and — stated fairly — the least proven at scale today. Judge it on pilots and references, not brochures.
The control question that separates the category
An agent that can send email, book jobs and chase money can also do those things wrong, at machine speed. The only acceptable default for a business's office is an approval law: drafts and actions are prepared automatically; nothing customer-facing or money-moving happens without the owner's yes. Any vendor whose demo skips the approval step is demoing the risky part as a feature.
Cost shapes, side by side
- Office manager: ~$3,500–$5,000/month loaded, plus hiring risk and single-point-of-failure risk.
- Managed AI back-office service: publicly listed around $499–$1,999/month.
- Platform software: hundreds per month per business, plus the salaries of everyone who operates it.
- AI teammate suite (Deck 72): pricing is set at launch; the model to expect is a platform fee plus per-outcome charges — hold us to publishing it plainly, the way this guide demands of others.
How to evaluate, in one afternoon
- List your office's ten recurring jobs. Ask each vendor, job by job: done by the system, assisted, or untouched?
- Ask what it can do without asking you. Write the answer down.
- Ask where the memory lives and what happens to it if you leave.
- Ask for per-job profitability: show me one job's true margin.
- Check availability without embarrassment — pilot, waitlist, or shipping. (This guide marks its own pick as pilot-stage; demand the same candour everywhere.)