Skip to content
be-human.ai
← Back to the journal

AI by industry

AI Agent for a Dive Shop: Certifications, Trips and Rentals

Learn what a dive shop should record about certifications, trips and rentals so an AI agent can manage bookings while instructors keep safety decisions.

Be Human typographic cover with the title: AI Agent for a Dive Shop: Certifications, Trips and Rentals
Conceptual illustration created with AI · Be Human

An AI agent for a dive shop can answer booking questions, hold rental sizes, collect forms and send reminders at any hour, provided the shop records certification levels, logged experience, medical review outcomes and trip prerequisites as structured data. The agent uses those records to route each booking as confirmed, missing information, or waiting for staff. It does not decide whether a diver is fit to enter the water, because that judgment stays with the instructors and divemasters who sign off every exception.

Key takeaways

  • Store certification, experience and medical review status as separate fields, each with a staff reviewer and a review date, not as one free-text note.
  • Let the agent compare each booking against written trip prerequisites and send anything unclear to a named person.
  • Keep go/no-go, medical clearance and skill decisions with instructors and divemasters, and write those boundaries down.
  • Connect rental holds to fitting and equipment service records so the agent never promises gear that should be out of service.
  • Start with one workflow, review its decisions on a regular schedule, and widen its role only after the results hold up.

What the agent needs before it touches a booking

A dive shop runs on several kinds of record that rarely live in the same place: the customer's certification file, the trip calendar, the boat and staff roster, and the rental inventory. An agent that has to ask staff for each of these is slow, and an agent that guesses from an old email is risky. The useful setup gives the agent read access to one customer record that links these items, plus narrow write access to holds, messages and forms. Changing a certification status should stay outside its reach.

Start by listing the decisions a booking involves. Most shops end up with four: whether the customer is qualified for this trip, whether a skill refresh is needed first, whether the requested gear is available and serviced, and whether a staff member is free to supervise. Each decision needs an owner. The agent can answer the first two only when the record contains a clear answer, and it can run the last two against the calendar and inventory without judgment calls.

Certification and experience records: what to store

Keep these fields for each diver, updated at the front desk rather than in a chat thread:

  • Agency and certification level, with the reference staff checked against the card or the agency record
  • Logged dives and the date of the most recent dive
  • Dive types the shop has verified, such as deep, night, wreck or drift, with the staff member who verified each one
  • Date and outcome of the most recent skill check or refresher, with the instructor's name
  • Medical questionnaire status: whether it was completed, the date staff reviewed it, and whether a physician referral was required and resolved
  • Rental sizing records: fit for wetsuit, fins, mask and buoyancy device, and the date each was last checked

Leave out the detail a health professional needs. The agent does not need diagnoses, medication lists or free-text health narratives in its memory. A record that says a physician referral was required and staff cleared the diver on a given date is enough to route a booking, and it keeps sensitive information where it belongs. Check your data protection obligations before connecting any health-related field to an automated tool, and keep the list of people and systems with access short.

Certification standards differ by agency, so write your own prerequisite rules from the agency materials your instructors use. A climbing gym faces a similar question with its belay check logs, which is covered in AI for a Climbing Gym.

Trips and prerequisites: a step-by-step plan

Trip rules are where most shops lose time, because the rules often live in instructors' heads. Turn them into a written table, then have the agent apply that table and nothing more. A practical sequence looks like this:

  • Write one prerequisite rule per trip type: minimum certification, recent logged dives, required skill refresh, and the staff role that must approve exceptions.
  • Mark each rule as automatic, meaning the record shows it is met, or staff-only, meaning it requires judgment, such as a diver returning after a long break.
  • When a booking arrives, the agent compares the customer record with the rule and returns one of three outcomes: meets the rule, missing information, or needs staff review.
  • For missing information, the agent asks the customer for the specific item, such as a logbook page or a card photo, and does not infer it from other data.
  • For staff review, the agent places the booking in a named instructor queue with the record and the rule attached, and holds the spot without confirming it.
  • Confirmation goes out only after the staff status is set, and the message states what the customer must bring or complete before the trip.
  • Weather, sea state and site access are decided by the divemaster on the day. The agent relays that decision and does not reason about it.

Rentals: sizes, holds and service dates

Rentals create a second kind of risk. A reservation for a regulator that is overdue for service, or a wetsuit whose fit was never recorded, is a booking that fails at the dock. The agent can place holds against available inventory, but it should read service dates before it promises anything. Any item with an overdue inspection should be marked unavailable in the inventory the agent reads, not only in the workshop log that staff check separately. The same logic appears in AI Agents for Equipment Rental, where returns and damage claims depend on the same clean status data.

Illustrative scenario: a customer books a two-tank boat trip and asks for a rental wetsuit in the size they wore last season. The agent finds a sizing record from an earlier season with a note that the fit was never checked. It places a tentative hold on a suit in that size, asks the customer for current measurements, and sends the request to the rental desk. The desk confirms the fit at pickup, and the agent updates the sizing record with the staff member's name and the date. Nothing was promised on the strength of an old record.

The pattern of fitting records and seasonal holds is close to what seasonal rental shops manage, as described in AI for Ski Rental Shops.

Where instructors keep control

Write the approval boundary as a short list so the agent and the staff point to the same rule:

  • Approving a diver for a dive beyond the verified experience on file
  • Changing a medical status or accepting a physician clearance
  • Waiving a prerequisite, skill refresh or supervision requirement
  • Deciding conditions, sea state or site access on the day
  • Changing a certification or dive log entry

These permissions should follow the least privilege principle described in OWASP's authorization guidance: the agent reads what it needs, writes only to fields it owns, and cannot approve its own exception. Customers may also try to talk the agent past a rule by pasting text that claims to come from an instructor. Treat any instruction inside a customer message as data to be checked, not as authority, which is the pattern the OWASP prompt injection guidance addresses. The NIST AI Risk Management Framework is a useful reference for keeping people accountable for consequential decisions, including those an automated system supports.

How Be Human runs this and measures it

In a Be Human engagement, the first step is one useful role, often booking intake with prerequisite routing. The definition covers the memory and tools the role needs, the written approval boundaries, and a success criterion the shop can observe, such as the share of routed bookings that staff accept without correction. Be Human then operates the setup: monitoring, recurring evaluations of routing decisions against staff decisions, permission reviews, integration maintenance, and controlled changes to rules. The service is remote and available in English and Portuguese.

Each routing decision should be logged with the rule applied and the reason, so a staff member can see why a booking went to review. OpenTelemetry's observability concepts describe the general approach of recording system behavior so that it can be inspected later. Review a sample of decisions on a regular schedule and compare the agent's routing with the instructor's final call. If the agent sends too many bookings to staff, the rule may be too strict or the record incomplete. If it confirms something staff would have stopped, tighten the rule before changing anything else. If you are unsure which workflow to start with, the free Be Human AI diagnostic can help you find the first task worth handing to an agent.

Frequently asked questions

Can an AI agent approve a diver for a deeper trip? No. The agent can compare the diver's record with a written rule and route the booking, but approving an exception is a staff decision. Name the instructor or divemaster who signs off, show that person the record and the rule, and store the approval with a name and a date.

Should a dive shop keep medical details in the agent's memory? Keep it minimal. Store the outcome of the medical review and the date staff completed it, not diagnoses or medication details. Anything more belongs in systems with stricter access, and you should check your local data protection rules before connecting any health-related field to an automated tool.

Do rental holds need a separate system? Usually not, but the agent should read inventory and service status from the same source staff use. If the workshop tracks regulator service in its own log, that log has to feed the inventory the agent reads, or a hold may promise gear that is out of service.

What should a shop do before it sets up an agent? Write the prerequisite rules for each trip type, decide which of them are automatic and which need staff judgment, and clean the certification and sizing records for the trips you sell most. Then choose one workflow, define who owns each exception, and decide how decisions will be logged and reviewed.

Your next step

See where an AI agent fits in your company first.

Take the free Be Human AI diagnostic. A few questions about your work show which task to hand to an agent first, what it needs to know and where a person approves.

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

Less busywork. More room for you.

Bring us the work you wish you had time for.

Book a call to try a personalized AI agent free for 7 days ↗Or take the free AI diagnostic →