How to read buying intent from conversations with Audience Stages

You read buying intent by deciding what it looks like before the conversations arrive. In Audience, you write plain-language criteria for each stage, a Digital Mind can place contacts as they reply, and a person reviews and corrects the result. Treat every placement as an assessment from the conversation, not a verified fact.

Why do conversations show more than a lead form?

A lead form records the fields you thought to ask for, at the moment someone agrees to fill them in. It rarely says why they asked, what they have already tried or how soon they need an answer.

A conversation holds that context in the visitor's own words. It also leaves a record a person can read: the message that made someone look interested is there to inspect and, if the reading was wrong, to correct. A number produced out of sight offers neither. The case holds only if the criteria are written down and the results stay reviewable. Lead forms vs conversations covers the comparison.

What is Audience, and where do Stages fit?

Audience is the part of Plandora that connects each conversation with the person behind it. Three areas matter for reading intent.

There is no hidden number to decode: a contact sits in a named stage whose criteria you wrote.

Availability varies. Plandora's public pages say Audience may need a guided rollout, so check what your plan or rollout includes before you build a process around it.

How do Stages work?

Start with the journey, then the settings. Setup begins with a choice: describe your journey or keep the stages you already have. Then it walks through these steps.

1. Review each stage's name, order and Move here when… criteria.
2. Choose the default stage, where new contacts start, and the won stage.
3. Set each stage to Automatic or Manual.
4. Check linked automations, and where contacts from removed stages should go.
5. Select Activate Stages.

Nothing changes before activation.

An automatic stage lets the Mind move a contact as it replies, when a reply gives enough evidence. A manual stage waits for your team. The won stage, the one that means a customer, is always manual: a visitor saying they will buy is not a purchase, and only your team can confirm one. Bookings follow the same logic. A click on a booking link shows interest, and your team confirms the outcome in Stages.

Two controls keep you in charge. Moving a contact yourself pauses automatic updates for that contact, which then shows Managed by you; Resume automatic updates lets future replies move them again, without re-sorting old messages. And Sort existing contacts reads recent messages and moves contacts when a stage clearly fits. It skips contacts you manage and those in manual stages, uses AI credits (an estimate appears first) and does not trigger automations. Other stage changes can, so review linked automations before you edit the journey.

How do Data Collections add detail?

A data collection is a set of fields your Mind fills in during chats, like a form it completes for you. Define a field such as Timeline or Service interest, pick a type that fits the answer, then tag it in the Flow step of your Playbook where the Mind should ask for it. Creating a field alone does not make the Mind ask.

The Mind records details the visitor clearly states, and the answers sit on the contact record beside the stage. You can correct them, show them as table columns and use them in groups. Where each kind of instruction belongs (knowledge, Playbook, fields, Automations) is mapped in the expertise operating system.

One boundary matters. The docs describe collected answers as contact details you can filter, group and export, and they never say those answers feed stage decisions. Write stage criteria that stand on their own in the chat, and use fields for the specifics your team needs when it follows up.

How do you write stage criteria in plain words?

Write each Move here when… line as something a person could check by reading the chat. Describe what the visitor said or did, not what they seem like. "Seems serious" cannot be checked. Has described a specific problem and asked about price, availability or how to start can.

A few habits hold up:

• One stage, one change in what the visitor has shown. If a line needs "and also", split the stage.
• Name the evidence: a stated problem, a stated deadline or a direct request, rather than tone.
• State the boundary. Say what does not qualify, such as a price query with no problem attached.

This illustrative journey for a generic service business shows the shape. The names and wording are examples, not defaults.

Early stages suit Automatic because a wrong placement costs little. Stages that trigger a message or someone's time suit Manual, since a stage change can start an automation.

What signals in a conversation point towards buying?

Signals worth writing into criteria:

• A specific problem described in detail, where a topic alone would be vague.
• A consequence or deadline: why it matters now, and what happens if nothing changes.
• Practical questions about price, availability, how to start or what happens afterwards.
• A question about fit for one particular situation, or a comparison with alternatives.
• A direct request for a quote, a call, a person or a date.

Counter-signals matter as much: general curiosity, requests for free material only, questions about someone else's situation, and a price question with no problem attached. None of these rules out future interest. They mean the evidence is not there yet.

What do older qualification frameworks add?

Two frameworks give useful vocabulary. BANT stands for Budget, Authority, Need and Timeframe. IBM is usually credited with it, though a primary account of its origin could not be confirmed here; what is documented is IBM's own use. An IBM PartnerWorld form for government sales opportunities (opens in a new tab)^[1] (template v1.1, file dated June 2012) says: "A validated opportunity is one where you have made at least one sales call and know the customer's Budget, Authority, Need and Timeframe." The same form says marketing campaign and seminar leads do not count as validated opportunities.

That standard was written for a salesperson on a call. In a first chat with a stranger, a direct budget question can read as a gate.

SPIN is the other. Huthwaite International (opens in a new tab)^[4], the firm behind it, describes SPIN Selling as a research-based method formed from more than 35,000 sales calls (a vendor-reported figure). Its four question types are Situation, Problem, Implication and Need-payoff. For criteria, the lesson is that a visitor who explains the implication of a problem, what it costs to leave it alone, has shown more than one who only states a situation.

Neither framework was designed for an AI chat, and neither proves what predicts a sale in your business. Use them as vocabulary, then test your criteria against your own conversations.

What can the system know, and what can't it?

A stage is a system-assessed placement based on what was said. It is an opinion about a conversation, and opinions can be wrong: visitors exaggerate, ask on behalf of someone else, joke or get misread. The Insights measure Answered carries the same caution: it is a system assessment, not proof the visitor was satisfied. Measuring AI conversations covers how to read such measures.

The system cannot know that someone bought. That is why the won stage is manual. A contact in Qualified is a candidate for follow-up, not a forecast, and a stage move is not proof that your Digital Mind caused a sale, so do not report purchases or attribution from stage moves alone. It also cannot identify anonymous visitors: Plandora's FAQ says they stay separate until there is a permitted way to identify them.

Recorded assessments can carry a data-protection duty too. In the UK, Article 5(1)(d) of the UK GDPR (opens in a new tab)^[2] requires personal data to be "accurate and, where necessary, kept up to date". The ICO's guidance on accuracy (opens in a new tab)^[3] adds that a record of an opinion should make clear it is an opinion and, where appropriate, whose. That is a UK rule as read on 22 October 2026, and the ICO marks its page as under review after the Data (Use and Access) Act. Other jurisdictions differ, and the business using the Mind is responsible for notices, permissions and lawful use of data. This is not legal advice; ask a qualified adviser. Keep a person involved in financial, legal, medical and other consequential decisions.

How do you review and correct stages?

1. Filter Inbox by an automatic stage and open a few contacts. Read the conversation first, then the contact details.
2. Check the stage history on the contact's Activity tab. It can show who made each move (you, the Mind, an agent, an automation or the system), a short explanation and, where one was recorded, an excerpt from the chat. A move of yours that overrides the latest automatic one is marked as corrected by you.
3. If a placement is wrong, move the contact yourself. Their automatic updates pause, so the next reply does not overwrite your correction.
4. Select Resume automatic updates when you want future replies to move them again.
5. If the same mistake keeps appearing, change the criteria rather than the contacts. After activation, ordinary stage edits save automatically, and Sort existing contacts can re-check eligible contacts against the new wording.

Connect an automation to a stage only after you have read enough placements to trust its criteria. A stage move is also separate from who handles the conversation: when a visitor needs a person, use take over from Inbox, and see human handoff for AI conversations for when to bring one in. Once stages are trustworthy, the next gain is replying quickly to the people who are ready, covered in speed to lead with a Digital Mind.