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

AI by industry

Superintelligence for Craft Distilleries: Tasting Bookings and Batch Records

A practical playbook for linking tasting bookings, batch records and compliance logs so distillery staff find current facts in one place.

Be Human typographic cover with the title: Superintelligence for Craft Distilleries: Tasting Bookings and Batch Records
Conceptual illustration created with AI · Be Human

Superintelligence for a craft distillery means AI agents that work from one shared, governed record of the facts the business depends on: which barrels sit in which batch, which tastings are booked and who is coming, and which compliance entries exist for each lot. The goal is that a tasting host, a production lead or an owner who asks what is true today gets the same answer from every channel. This playbook shows how to link those three records, where the agent's authority should stop, and what to check before handing any of it to software.

Key takeaways

  • Tasting bookings, batch records and compliance logs describe the same liquid from three angles, so they should point to shared identifiers instead of keeping copies of each other.
  • Staff lose more time to conflicting information than to missing information, so the first task is agreeing which record owns each fact.
  • An agent can read, match, draft and flag, but releases, regulatory submissions and guest exceptions stay with named people.
  • Every agent action should leave a trace a manager can review, including the records it read and the rule it applied.
  • Start with one narrow workflow, confirm that staff reach the current answer faster, and widen permissions only after a review.

Why tasting bookings and batch records drift apart

A tasting booking is a promise made to a guest: a date, a time, a party size, a flight of pours and sometimes a bottle the group can take home. A batch record is a production fact: the mash bill, the fill date, the barrel identifiers, the proof at bottling and the current release status. A compliance log holds the dated events that your licensing authority or your own quality program requires you to keep, such as transfers, sample pulls, label approvals, losses and corrective actions. In many small distilleries these three live in separate tools, or partly in a notebook beside the still.

The drift happens at the seams. A barrel is put on hold after a quality question, but the tasting team has already promised that exact expression to a group of eight. Nobody has been careless. The hold simply never reached the booking calendar. Later, when someone asks whether a pour came from a released lot, the answer requires opening three systems and two conversations. The same pattern appears in other hospitality businesses, which is why the seams between events and production are worth the same attention. Superintelligence Prep for Breweries examines a similar split between taproom events and keg orders.

Which record owns which fact

Before any automation, write down an ownership table in plain words. Keep it short enough to fit on one page, and name the person accountable for keeping each part correct. Unclear ownership is the most common reason linked systems produce confident wrong answers.

  • Batch records own production facts: recipe version, fill date, barrel identifiers, bottling volume and release status.
  • Compliance logs own dated events and approvals: transfers, sample pulls, label approvals, losses and corrective actions.
  • Tasting bookings own guest facts: date, party size, requested expressions, age check status, dietary notes and contact preferences.
  • A shared lot code connects all three. A booking refers to lot codes, never to a free-text product name such as a nickname that changes from season to season.

The working rule is that no record copies a fact it does not own. A booking can show that Lot 14 is on the tasting menu, but the release status must come from the batch record and the approval history from the compliance log. When the owning record changes, the other records read the change instead of keeping their own note. Status words also need to be fixed in advance. Choose a short set, such as available, held, released, sold out and pending review, and use exactly those words in every system and every staff conversation.

A step-by-step plan to link the three records

  • Step 1: List the ten questions your staff ask most often in a normal week, such as whether a pour is available, whether a lot is released or who approved the last label change. Keep the list to questions with factual answers.
  • Step 2: For each question, name the one record that owns the answer and the person who keeps that record correct.
  • Step 3: Standardize identifiers. Every barrel, lot and bottling run needs one code that appears in all three records. Retire free-text product names in the booking tool.
  • Step 4: Define the status vocabulary in plain language and publish it to the tasting team and the production team at the same time.
  • Step 5: Give the agent read access to all three records and write access only to drafts, flags and booking notes. It should not be able to change a batch status or a compliance entry.
  • Step 6: Write the conflict rule. If a booking references a lot that is held or unreleased, the agent flags the booking, drafts a message for the host and waits for a person to choose a substitute.
  • Step 7: Log every read, draft and flag with a timestamp, the source record and the rule applied, so a manager can reconstruct any answer later.
  • Step 8: Review the log every week during the first month. Adjust the rules where the agent flagged noise, and where it missed something a veteran staff member already knew.

The last step matters more than it looks. Experienced staff often carry exceptions in their heads, such as a lot that is always poured last or a guest group that needs a specific glass. Those exceptions should be written into the rules deliberately, or they should be reviewed by a person each time.

What the agent may do and what stays with people

Write authority down as a list before the first deployment. An agent with vague permissions will eventually act on something it should only have described.

  • Allowed: answer staff questions from the linked records, draft tasting confirmations, flag bookings that reference held or unreleased lots, and summarize the day's open compliance items.
  • Allowed with review: propose a substitute expression for a guest, draft a corrective action entry, or prepare a transfer summary for a manager to sign.
  • Not allowed: change a batch status, sign or submit a compliance entry, waive an age check, promise a guest a specific release, or edit a historical record.

Access should follow least privilege, the idea that each identity receives only the permissions its task needs, and that sensitive actions are checked when they happen rather than assumed from a login. Treat any text inside a guest message, supplier email or uploaded document as untrusted input, since it can contain instructions written to change the agent's behavior. The NIST AI Risk Management Framework is a useful general reference for deciding who owns these risks inside your organization, and it does not require any particular software. Keep the agent away from anything that cannot be undone without a human decision.

Illustrative scenario: a Saturday tasting after a batch hold

Illustrative scenario, not a real business: a distillery has booked a group of eight for a Saturday tasting that features a single-barrel expression. On Thursday, the production lead places that barrel on hold after a routine check reads outside the expected range. The hold is recorded in the compliance log with a reference to the batch. The agent, which reads all three records, notices that the booking refers to the same barrel code and marks the booking as conflicting. It drafts a note for the tasting host listing two released alternatives, and a message for the guest, then stops. The host reviews the draft, the owner approves a substitution, and the booking record shows who approved it and when. The batch record is unchanged, and the compliance entry was never edited.

The scenario is useful because the failure it prevents is ordinary. The information existed. It sat in the wrong place at the wrong moment, and nobody was positioned to connect it before the guests arrived.

How Be Human's managed approach handles this

Be Human starts with one workflow, usually the link between tasting bookings and lot status, and defines the role, the memory it needs, the tools it can call and the approval boundaries before anything runs. Managed operations then cover monitoring of the flags the agent raises, recurring evaluations against real past questions from your staff, maintenance when a booking tool or batch system changes, periodic permission reviews, and controlled updates to the rules. Decisions about releases, filings and guest exceptions remain with your team. The aim is to keep the agent's memory and permissions visible and under human control, so that each answer can be traced to the record that supports it.

Where to start

Choose the one question your team asks most often that currently needs three systems to answer, and write down the record that should own each part of it. For many distilleries, that question involves whether a pour booked for next weekend is still available under the current lot status. Hospitality teams that already handle date holds and vendor timing will recognize the pattern, as seen in AI Agents for Wedding Venues. If you want help identifying the first task worth handing to an agent, take the free Be Human AI diagnostic.

Frequently asked questions

Is superintelligence for distilleries just a chatbot for tasting bookings? No, a chatbot that only answers guest questions does not solve the record problem. The useful part is an agent that reads the batch and compliance records, checks them against each booking, and flags conflicts for a person to resolve. Guest-facing messages can be part of the workflow, but they should be drafted from verified facts rather than invented from memory.

Can an agent submit compliance records to a regulator for us? It should not. Submissions, signatures and attestations belong to accountable people, and the agent's role is to prepare summaries, point to the entries that support them and flag anything missing. Your licensing authority defines what must be recorded and how long it is kept, so those requirements should come from your own compliance process, not from the agent.

Which system should be the source of truth for batch status? The batch record should own release status, and every other system should read it. Bookings and compliance logs can reference the status, but they should not hold their own copy that can drift. If your batch system cannot yet expose status cleanly, start with a shared lot list that production updates daily, and treat that list as the single owner until the integration is ready.

How do we know the agent is reading the right record? Require that every answer cite the lot code and the source record it used, and review a sample of those citations each week during the first month. Recurring evaluations on real past questions show whether answers stay correct after changes to the tools or the rules. If a citation points to the wrong record, the rule or the identifier mapping needs correction before the agent is trusted with more work.

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 →