A B2B sales team
Connect does the parts of B2B selling that decay when people are busy: recording what was said, keeping the pipeline honest, chasing the commitments and grounding the answers. It does not price, discount or negotiate — a commercial term Knowledge cannot support is refused and escalated, which is what makes the rest safe to run unattended.
What the pipeline actually holds#
- Opportunity
- Opened against a person and a company, with a value stored in minor units so nothing depends on a currency assumption.
set_valueandupdatechange it;moveapplies the stage rules. - Stage rules
moveis not a free-text field change. Stages have rules, which is what stops a pipeline where everything is at 90 per cent.- Review flags
raise_for_reviewmarks an opportunity a person must look at;clear_reviewends that. Useful for the deals that look fine in a report and are not.- Demos
request_demo,schedule_demo,complete_demo,cancel_demo— a lifecycle rather than a diary entry, so a demo that never happened is visible as such.- Follow-ups
- The commitments underneath all of it, each with a channel, a due time and a reason.
The commercial boundary#
safe_sales.py is the rule that Connect does not invent commercial terms. A price, an SLA or a warranty Knowledge does not support is refused and escalated to a person rather than improvised. On a sales team that fires several times a day, and it should: the alternative is an agent creating obligations that a salesperson then has to honour or withdraw.
Two things are correctly reserved to the platform operator rather than available inside a customer workspace: verifying payments, and the operator own commercial pricing. Those are the business of running the platform rather than of using Connect, and the separation is deliberate.
Where a salesperson gets the leverage#
Through the Assistant, which carries 66 tools. The read-only ones that matter on a sales desk are customer_360 for one person whole history, next_best_action, list_opportunities, list_followups, find_anything and activity. The writing ones include create_opportunity, move_opportunity, set_lifecycle_stage, create_followup, complete_followup and remember.
Each Assistant tab keeps its own conversation, follow-up chain, record, draft, files and scroll, which is how a person works three accounts in an afternoon without losing the thread on any of them. The screen context the model is given is checked against the database before it is trusted.
Feeding the top of the pipeline#
Prospecting starts from criteria and public research rather than a bought list. Evidence is attached to each claim, research spend is routed so a deeper pass runs only where it can change a decision, and identity resolution stops an existing customer being approached as a new prospect.
The rule that caps volume is that no email address is ever guessed. A prospect with no discoverable address stays a researched prospect with a different next action. Teams that measure activity by messages sent find this uncomfortable; teams that measure it by replies usually do not.
What this does not give a sales team#
- No dialler. Outbound calls are placed under the same gates as any outbound action, two phone follow-ups per tick, with a row more than 24 hours late closed as
missed:rather than rung. - No SMS sequences. Outbound SMS is provider-dependent and the live carrier carries none.
- No forecast model. Stage rules and review flags are what exist; nothing here predicts a close date, and this page will not imply otherwise.
- No call recordings for coaching. Recording is foundation and is not enabled on the live carrier. Transcripts and summaries exist and are searchable, which covers review but not listening.
Questions#
Can Connect run a full outbound sequence for the team?
It can create and chase follow-ups across email and phone under the channel autonomy rules, which is the shape most sequences actually take. What it will not do is send to an address nobody discovered, ignore a suppression, or put a price in a message that Knowledge cannot support.
Who sets the deal value?
A person, or the Assistant acting on a person instruction with a figure that came from somewhere real. Values are stored in minor units, which is worth knowing when a number looks a hundred times too large or too small. Connect does not derive a value from a conversation on its own.
How do we review what was said on a sales call?
The transcript and the summary on the call record, plus price_spoken if you are checking whether the voice priced anything. There is no audio to replay, because recording is not enabled on the live carrier — for review purposes the searchable transcript is usually the better artefact anyway.