Montana
Skills soon
A skill is a job Montana does the same way every time, without being asked in a conversation. You switch it on, it runs on a schedule or on a button, and it comes back with work done rather than a report about work not done.
Everything on this page is designed and specified, and none of it is running in production today. It is written down ahead of the build so the behaviour is agreed before the code exists. Nothing here is billable and nothing here will surprise you by switching itself on.
The idea
Montana chat is a conversation: you describe what you want and it answers. That is the right shape for a question and the wrong shape for a job you need doing every Monday. A skill is the job, written down once and given a name, so it runs identically every time and nobody has to remember how to ask.
Every skill below is described the same way: what it does, meaning the actions it actually takes, and what you get, meaning why it is worth switching on. If a skill cannot answer the second one it should not exist.
How a skill is shaped
| Setting | What it controls |
|---|---|
| Enabled | On or off, per workspace. Off by default for anything that writes. |
| Schedule | When it runs. Leave it empty and the skill only runs when somebody presses Run now. |
| Authority | Propose or act. See below. |
| Budget | A monthly ceiling in dollars. A skill that hits it stops until next month rather than running up a bill. |
| Scope | Which campaigns it may touch. A skill cannot reach outside the ones you name. |
| Config | Per-skill settings: how far back to look, how many items to return, which thresholds matter. |
Propose before act
This is the part that makes it safe to point at a live sales floor. Every skill starts on propose: it does the work and writes down what it thinks should happen, and a person accepts or rejects. Nothing changes until somebody says yes.
A skill earns act by having a track record. Once its proposals have been accepted consistently, an admin can promote it and it stops asking. You can demote it again in one click, every run is logged either way, and every action it took can be undone.
Filling the queue
Skills that decide who your reps talk to. The cheapest gains on most floors are here, because the same people making the same calls to a better-ordered list connect more.
Queue shaping
What it does. Reorders the dial queue before the floor starts. It ranks by the things that actually predict a connect on your own data: industries that pick up for you, company sizes that convert, the hour of the day that works for each area code, and how long it has been since the last attempt.
What you get. More conversations from the same reps and the same list. Nothing is bought and nobody works longer; the order just stops being arbitrary.
Best-hour routing
What it does. Holds each lead until its own best window rather than dialling it whenever it surfaces. It learns the curve per timezone and per industry from your own connect history, so a west coast prospect is not called at seven their time because a rep in Karachi had a gap.
What you get. The single largest connect-rate lever most floors never pull, and it costs nothing but ordering.
Lookalike sourcing
What it does. Takes the accounts that actually converted, works out what they have in common, and goes and finds more of them. It drafts a lead search from the pattern, shows you the pattern it found, and enrols the results into the campaign you choose once you approve the spend.
What you get. A list that improves every month instead of decaying. Your best customers become the template for the next batch rather than an anecdote.
Reactivation sweep
What it does. Goes back through leads closed months ago for reasons that expire: no budget this quarter, using a competitor, call me next year, the champion left. It re-enrols the ones whose reason has aged out, at the right moment, into a cadence written for people who have heard from you before.
What you get. Pipeline out of a list you already paid for. On most floors the dead pile is larger than the live one and nobody ever goes back to it.
Data hygiene
What it does. Finds the holes that cost you calls: no direct number, no named contact, no email, a company recorded twice under two spellings. It fills what it can from what was said on calls, merges the duplicates, and files a task to the account owner for the rest.
What you get. Fewer dead dials, and a CRM your reports can be trusted against. Every floor has this rot and nobody has ever had a spare afternoon for it.
Making the conversation better
Skills that work on what is said once somebody picks up.
Objection miner
What it does. Reads the week's reviewed calls, groups the objections reps ran into, ranks them by how often they came up and how badly they ended the call, and writes the top ones into the company brain with a real example and the answer that worked best.
What you get. You find out what is actually being said on your phones without listening to forty recordings, and the answer that works spreads to everyone instead of living in one rep's head.
Talk track experiments
What it does. Runs two openers against the same list at the same time, splits them evenly across reps and hours so the comparison is fair, and calls the result once there is enough data to mean anything. Then it writes the winner into the brain and retires the loser.
What you get. Script decisions settled by evidence in a fortnight rather than by whoever argued hardest in the Monday meeting.
Rubric tuning
What it does. Compares what your scoring rubric rewards against what actually produced meetings, and proposes changes where the two disagree. If calls that score badly keep booking, the rubric is measuring the wrong thing and it says so.
What you get. Coaching aimed at what closes rather than at what somebody guessed would close when the rubric was written.
Top rep interviewer
What it does. Takes the calls that went best, works out what that rep did differently, and turns it into coaching for everyone else: the opener they use, how they handle the price question, where they ask for the meeting, how long they let a silence run.
What you get. Your best rep's method written down, without them having to teach and without anyone sitting in on their calls. It is also the fastest way to shorten ramp for a new hire.
Nothing falling through
Skills that chase the things a busy floor drops.
Overdue escalation
What it does. Watches follow-ups past their date, callbacks that were promised and missed, and deals that have gone quiet against their usual rhythm. It raises them to the owner first and to the manager if they are still sitting there.
What you get. The Friday afternoon job every sales manager does by hand, done on Friday morning, on every account rather than the ones they remembered.
No-show rescue
What it does. Notices a booked meeting that nobody attended and acts within the hour, while it is still a scheduling problem rather than a lost deal. It drafts the reschedule from the original thread and puts the account back in the queue at the right time.
What you get. A meaningful slice of your no-shows recovered. The difference between reaching out the same day and the following week is most of the outcome.
Stalled deal detector
What it does. Learns how long a deal normally sits at each stage for your business, then flags the ones sitting too long with what it thinks is blocking them, taken from the last conversation rather than from the stage name.
What you get. Forecast you can believe, and a prompt to act while the deal is still warm instead of at the end of the quarter.
Keeping you safe and solvent
Consent and DNC sweep
What it does. Re-checks the list against your do-not-call entries and messaging opt-outs, catches numbers that entered through an import after somebody had already refused, and removes them before a rep dials.
What you get. The complaint that never happens. One repeated call to somebody who asked you to stop costs far more than the whole exercise.
Number health watch
What it does. Watches your connect rate per number for the drop that means a carrier has started marking it as spam, and tells you which number and when it started rather than leaving you to notice a bad month.
What you get. You find out your caller ID is burned in days instead of quarters, which is the difference between rotating one number and rebuilding a floor's output.
Spend watchdog
What it does. Watches minutes, AI usage and lead-sourcing spend against where you are in the month, and warns before a ceiling bites rather than when a rep cannot dial. It flags the unusual too: a sudden run of expensive international destinations, a single account burning a third of the pool.
What you get. No surprise invoice and no floor stopped mid-morning because a limit was reached quietly.
The manager's Monday
Weekly briefing
What it does. Writes the update a sales manager writes by hand: what changed this week, which campaign moved, who improved and who slipped, the objection that appeared out of nowhere, and the three accounts worth personal attention. Sourced from calls, not from activity counts.
What you get. An hour back every week, and a version that reads the calls rather than the dashboard, which is where the actual news is.
Writing your own
The first release ships the skills above. A workspace will later be able to define its own: a name, a description of the job in plain words, a schedule and an authority level, the same way the built-in ones are shaped. That is deliberately the same idea as skills in Claude, so anyone who has written one there can write one here.