Two pharmacies are waiting for an answer, a shop has just received its first order and you have an appointment at 2 pm. Sorting customers by revenue is not enough to organise the day. You need a reason to visit each account and a schedule that respects the appointments.
AI can help review priorities from a supplied customer record. Travel times and appointment windows still need a routing tool and confirmation. The examples here are fictional and apply to reps visiting shops and pharmacies.
Separate visit priority from route calculation
A language model can group open commitments, suggest questions and flag missing information. It should not invent travel times or buyer availability. A routing planner takes addresses, travel times, visit durations and time windows to calculate a feasible sequence. A high commercial priority does not make a late arrival possible.
Google’s vehicle routing example with time windows illustrates this distinction: the travel-time matrix and allowed visit windows are inputs to the calculation. The quality of the result depends on those inputs.
- Prioritise
Find the visit purpose and outstanding commitments in the CRM.
- Calculate
Give the planner travel times, visit windows and durations.
- Confirm
Check the buyer, access and buffer time before departure.
Confirm unknown availability. Never invent travel times.
Prepare a short list of candidate visits
| Data | Useful example | If missing |
|---|---|---|
| Outlet | CRM ID and checked address | Confirm the address; do not choose a namesake |
| Visit purpose | Check shelf setup after the first delivery | Clarify the objective before reserving time |
| Open commitment | Quote response promised before 12 September | Read the latest visit report |
| Availability | Buyer present 10 am–noon, by appointment | Confirm; shop opening hours are insufficient |
| Duration and access | 35 minutes on site, parking five minutes away | Set an explicit assumption to verify |
| Dated history | Last visit 18 August; next action 9 September | Flag the gap without automatically lowering priority |
Limit the extract to accounts in the territory under consideration and the information needed for the decision. An entire address book will not improve a decision about eight visits. Keep each outlet’s ID in the proposal so you can find the exact record.
Use rules the team can explain
Start with confirmed appointments. Then review commitments nearing their deadline, launches that need support and overdue visits. Commercial potential belongs in this judgement alongside the reliability of the data used to estimate it.
- →Fixed appointment: preserve it as a constraint unless the customer agrees to move it.
- →Outstanding promise: establish whether it requires a visit. Sending a price list may avoid a 40 km journey.
- →New shop: allow for launch support even without an order history.
- →Less active customer: check closures, disputes and seasonality before assuming the account is at risk.
A score of “87 out of 100” is of little use if nobody knows what increases it. For a first test, request a dated reason and an expected action for each visit. If you use a score, document the rules; a number does not become a probability of an order simply because it looks precise.
Make the day fit, then check each appointment
From 8:30 am to 5 pm, you have 510 minutes. Here is a fictional budget for six 35-minute visits. Travel includes departure and return; the planned durations also allow for parking.
| Item | Minutes |
|---|---|
| Six visits × 35 minutes | 210 |
| Planned travel | 145 |
| Lunch | 45 |
| Preparation and reports | 30 |
| Buffer: 510 − 210 − 145 − 45 − 30 | 80 |
A seventh 35-minute visit with a 25-minute detour would use another 60 minutes. Only 20 minutes of buffer would remain, before any additional admin. The visit count may fit on paper while the risk of lateness increases.
A correct total can hide an impossible appointment
You leave a shop at 11:30 am. The next journey takes 35 minutes, but the buyer is only available until noon. Arriving at 12:05 pm is already too late. Change the order, agree another slot or postpone the visit. Having 80 spare minutes elsewhere in the day does not resolve this conflict.
Available time510 min
- Six visits210 min
- Travel145 min
- Lunch45 min
- Preparation and notes30 min
- Buffer for delays80 min
This total does not prove that every appointment fits. Check arrival and finish times too.
Validate the plan and keep a fallback
Give checked addresses, appointment windows and visit durations to your routing tool. Review the arrival and finish time of every visit, then confirm appointments. Show traffic assumptions and optional stops in the plan.
If a customer cancels, choose a prepared fallback visit or use the time for a remote action. Recalculate while parked. Adding the nearest shop without checking whether its buyer is available can waste time and disrupt later appointments.
Measure useful visits, not just visit count
For two weeks, record appointments kept, kilometres travelled, commitments resolved, late arrivals and visits with no relevant contact available. Compare days with similar territories and constraints. Five visits that move a decision forward may be preferable to eight calls with no outcome; revenue alone does not capture that distinction.
Review your route planning
Bring a real territory, your appointments and the information available in your customer records to prepare the demonstration.
Read the practical field-sales AI guide or prepare voice notes after each visit. For more detail, our guides to organising a sales route (French) and territory workload (French) cover the planning fundamentals.
By Anatole Briant
By Anatole Briant
By Pierre Lerebours
COO of Salesia
Pierre Lerebours is COO of Salesia. He is your contact for implementation, training and ongoing support.