The phone agent does not belong in your technology budget. It belongs next to whatever it replaced.
That sounds like an accounting quibble and it is not. A line item buried in a general software bucket, sitting between your POS subscription and your scheduling app, is a line nobody can evaluate. A line coded next to the answering service it displaced or the phone position it shrank is a line that answers its own question every time someone reviews the P&L.
Put it next to the thing it displaces
Most restaurants land in one of three situations, and each suggests a different home for the cost.
If you were paying an answering service, the swap is direct. The new line goes exactly where the old one was, the comparison is visible on the same row of last year's statement, and the ongoing review is a single subtraction. Nothing else you can do to the chart of accounts makes the decision as easy to audit.
If you were staffing a phone position, the line belongs adjacent to labor, coded as a service cost rather than payroll but reviewed alongside it. The relationship you want visible is between the subscription and the hours it took off the schedule, and converting answered calls into hours sets out how to size that honestly, including why the honest figure is usually smaller than the enthusiastic one.
If neither applies and you were simply missing calls, the line belongs with your order-channel costs, next to delivery commissions and online ordering fees. That is the correct company for it, because it is doing the same job those lines do: bringing in off-premise orders at a cost per order. Framed that way it usually looks cheap, which is the comparison in voice AI versus delivery app commissions.
The fixed part and the moving part
The base plan is the easy half. Plans start at $250 a month and a fixed subscription is the most budgetable expense there is.
The moving half depends entirely on what your plan meters. Some pricing shapes include an allowance of minutes or calls and bill above it; others do not meter at all. Your bill's variability follows directly from that structure, and it is worth knowing which shape you are on before you build a twelve-month budget. Both shapes are modeled out in per-minute versus per-call pricing math, and pricing models covers the category more broadly.
If your plan meters, your bill has your seasonality in it. That is not a problem, it is a fact to budget around:
- A pizza or wings concept should expect its heaviest metered months around football season and the first weekend of March Madness, not evenly across the year.
- A catering-heavy operation will see November and December run well above the rest of the calendar, sometimes by half again.
- A patio-dependent restaurant will see summer volume that its winter average never predicts.
Budget the twelve months unevenly. An even twelfth of a projected annual figure will be wrong every single month, low in your peaks and high in your slow season, which makes variance reports useless and trains everyone to ignore them.
The shortcut, if you do not want to build a monthly curve, is to shape the line the way you already shape your food cost budget. Whatever seasonality index you use for cost of goods is close enough for call volume at most concepts, because both track the same underlying thing, which is how many people ordered from you that month. That gets you a defensible twelve-month plan in about a quarter of an hour.
First-quarter costs that do not repeat
Three things land early and then stop, and they should be budgeted once rather than annualized.
Carrier charges for porting a number or setting up forwarding, which vary by provider and are usually modest. Any integration fee if your POS reaches the agent through a middleware layer rather than directly, which depends on your system: Square, Clover and OrderCounter connect directly, while Toast, Lightspeed, TouchBistro, SpotOn, Aloha, Revel, PAR Brink, Micros and others go through Deliverect. And staff time, which is the one people forget because it never appears on an invoice.
That staff time is real. Someone has to verify the menu reads correctly, test the odd modifier combinations, and listen to a sample of calls for the first few weeks. Price it at a manager's loaded rate for a handful of hours, put it in month one, and do not carry it forward. Setup itself is typically under 24 hours, so this is a small project by restaurant-technology standards, but budget a manager's attention for the first month and it will go better than if you do not.
What stays in the budget forever
After the first quarter, the ongoing cost is short:
- The subscription, fixed and predictable.
- Metered usage if your plan has it, which tracks call volume and therefore tracks your season.
- A small recurring allowance for menu maintenance, concentrated around seasonal changes and limited-time offers, since a menu that has drifted out of sync is the most common cause of a bad month.
Nothing else recurs. This is a cheap line to run once it is running, and the maintenance is mostly attention rather than money. Seasonal menu updates covers what that attention consists of.
Reviewing the line
Twice, on a schedule, and against something specific.
At ninety days, check the two claims the purchase rested on. Did recovered orders show up in the numbers, measured against the baseline you took before go-live. Did the answering-service invoice actually stop, or did it quietly keep renewing because nobody cancelled it. That second one happens more than you would think.
Annually, check two different things. Whether your call volume has grown enough that a different plan tier is now cheaper than your overages, which is a question for your vendor and a fifteen-minute conversation. And whether your agreement carries an auto-renewal window or a price escalator, because both are ordinary contract terms and both are easier to handle sixty days before they trigger than sixty days after. Contract terms to avoid lists the ones worth reading closely.
If you run more than one location, resist the urge to hold this centrally. Push the cost to each store's P&L in proportion to its call volume, so the GM whose phone rings eleven hundred times a month sees a bigger number than the GM whose phone rings four hundred times. The central budget should carry the integration and reporting layer and nothing else. A cost that lands where the volume is generated is a cost somebody manages.