“Who approved that email?”
An AI draft reached a key account with the wrong price. Nobody checked it before it left.
ConvOps runs your sales playbook as governed AI agent workflows. Research lands as structure. A gate holds every send until it is approved. Follow-ups start from the meeting.
chatgpt · claude · nothing sends without approval · one motion for every rep
Northwind Logistics
A ConvOps workflow every rep's AI agent follows: research, scoring, a drafted message, then a gate that holds the send until it is approved.
ConvOps is the operations layer for AI agents. It holds your team's process as workflows with gates, a shared memory, scheduled and unattended runs, programs, a plan board, metrics and a full audit trail. It does not sell or host models.
key facts · october 2026
best for
not for
updated
AI writes outreach in seconds. Whether it is on message, accurate or checked depends on the rep.
“Who approved that email?”
An AI draft reached a key account with the wrong price. Nobody checked it before it left.
“Where did that decision go?”
The pilot scope agreed on the call lives in one rep's notes. The proposal never saw it.
“What happens when someone leaves?”
Their accounts, their context and their way of selling leave with them.
A real account, run through the Sales Outreach pack from ChatGPT. Every rep, the same motion.
You name the prospect and ask for an intro. ConvOps creates the task and attaches Prospect Research. The motion starts the same way for every rep.
Company, contacts and buying signals become fields on the task, deduplicated. The next step and the next rep read the same record.
Before drafting, the AI recalls the team's rulings: what logistics buyers respond to, and what never to say. Every draft starts from the same lessons.
askintro to a logistics ops lead
The draft waits at the approval gate. You read it, edit it or send it back. Outbound leaves only once the send is approved.
Send the intro to Dana Ruiz?
Subject: On-time rates at Northwind · 92 words
After the call, notes, commitments and next steps land on the same thread of work. The follow-up is drafted from what was actually agreed.
Thanks for the time today, Dana
The proposal runs in the same line of work. It starts from the research, the call and the decisions that led here. Nobody types the account again.
Pilot proposal: Lisbon dispatch
Five steps from first look to proposal. A gate holds every send until it is approved.
Companies and contacts land as structured, deduplicated tasks.
Written against your positioning, inside the process.
Every outbound message waits at the gate until it is approved.
needs approvalNotes, commitments and next steps on the same thread.
Inherits the research, the call and the decisions.
Labelled examples from different kinds of selling.
An SDR's agent finds one specific reason to write, such as a new job post, drafts 80 words with one ask, and waits at the send gate. The rep approves or sends it back.
Outreach Message send gate
A key account team runs a monthly plan for one banking client. Any message that mentions pricing or terms waits at a legal gate before the account lead approves the send.
Monthly schedule, two approval gates
Each reseller gets its own task on the same workflow: research, a tailored pitch outline and a cover note. The partner manager approves each pack before it goes out.
One workflow, one task per partner
A bid team splits an insurer's 120-question RFP into child tasks by section. The parent bid waits until every section is completed, and the bid lead approves the final response.
Child tasks and a final approval
Before each discovery call, the agent pulls the account's recent news, open roles and likely priorities into a one-page brief. There is no approval gate: the rep reads it before the call.
Research step, no approval
The rep pastes the call notes. Meeting Follow-Up drafts the recap with each agreed action, owner and date, and Send Review catches a wrong name before the rep approves.
Send Review approval
A freight forwarder's sales lead approves lanes and rates for a tender. The proposal draft is checked line by line against that approval before it reaches the client-ready gate.
Scope & Pricing approval
On the first weekday of each month, a schedule starts a review of the month's closed deals and writes the patterns into a short report. There is no approval gate: the sales lead reads it in the next team meeting.
Monthly schedule, no approval
The next rep and the next agent start with every lesson the team already paid for.
Logistics buyers care about on-time rates. Lead with that.
what wins
No pricing in a first email. Decided once, recalled before every draft.
decided once
Research, calls and decisions on one thread per account.
one thread
Case studies and pricing sheets, so no draft starts with a search.
where it lives
Prospecting is weekly work. ConvOps keeps it on a rhythm and shows when it slips.
Put research on a schedule: one segment every Monday, in your time zone, with a preview of the next five runs. Group the schedules into a program with an owner, and its verdict turns late or silent when a week is missed.
Plan follow-ups and proposals on Plan. A planned task can auto-run on its day at a set time in the task's time zone, and the send still waits at its gate.
Define the numbers outreach must report in Metrics, such as replies and meetings booked, and bind them at the step that logs the outcome. Every value carries who recorded it and at which step, and ConvOps records the cost and outcome of each run by itself.
A gate holds the send until an explicit approval is given, recorded with the account that gave it.
Prospect Research, Outreach Message, Meeting Follow-Up and Proposal. Inspect every step and gate before you install.
Change a step in the template and the next run, for every rep, uses it. Runs already in progress keep going.
Follow-ups and proposals run in the same task lineage, and decisions recorded on steps become memories the next run recalls.
Keep the sequencer and the CRM. ConvOps runs the researched work in between.
| ConvOps | A sequencer | An n8n or Zapier flow | A chat prompt | |
|---|---|---|---|---|
| What it does | Runs researched, one-to-one work as steps | Sends cadences to lists | Moves data between apps on a trigger | Writes one message on request |
| Before a send | A gate holds it until it is approved | Optional manual tasks | An approval step where the tool offers one | You copy and send it yourself |
| Same playbook for every rep | One stored workflow, edited once | Shared sequence templates | One flow per use | Each rep's own prompt |
| Research behind each name | A source for every match | Imported lists | Whatever the connected app holds | Whatever the model recalls |
| Record | Every step and approval on the task | Activity log per contact | Execution logs | The chat history |
Connect the AI tool your team already uses, install the workflows, and run one real piece of work through them.
Install the Sales Outreach pack
Point Prospect Research at your segment: it lands as structured tasks, not a doc
Draft outreach with the approval gate on: nothing sends until the send is approved
After the first meeting, the follow-up workflow already has the context
ConvOps is the operations layer for AI agents. It holds your team's process as workflows with gates, a shared memory, scheduled and unattended runs, programs, a plan board, metrics and a full audit trail. It does not sell or host models.
In ConvOps, the sales use case runs prospect research, outreach, meeting follow-ups and proposals as workflows every rep's AI agent follows. Research lands as a qualified list with a source for each account, messages are drafted on a researched hook, and a gate holds every send and every price until it is approved. What one step learns becomes the next step's input.
Yes. In ConvOps, the Outreach Message and Meeting Follow-Up templates put an approval gate on the send. A gate holds the step until an explicit approval is given, recorded with the account that gave it, so the agent drafts and the run waits. Remove the gate only when you decide the drafts no longer need a read before they go out.
The ones you define, then the run numbers. In ConvOps you define the metrics outreach must report, such as replies or meetings booked, and bind them at the step that logs the outcome, where the agent or a rep records them. Every value carries who recorded it and at which step, with evidence as a link or a note. ConvOps also records the cost, duration, tokens and outcome of every run by itself.
The playbook is the workflow. In ConvOps, every rep's AI agent receives the same steps from the same stored workflow, one at a time, with the same gates. Edit a step once, in conversation, and every run that starts afterwards follows the new version, for every rep and every AI client. The playbook stops being a document people half remember.
No. ConvOps governs the work around them. One way to split the work: the sequencer sends high-volume cadences, the CRM holds the account record, and ConvOps runs the researched, one-to-one work with a gate before each send. ConvOps does not send email itself; the agent sends through the tools its environment provides, or hands you the message.
Yes. In ConvOps, a weekly schedule can start a Prospect Research run for a segment, and a program groups those schedules under one owner with a health verdict. Plan shows the week ahead, and a planned task can auto-run on its day at a set time in the task's time zone. Sends still wait at their gate until they are approved.