Sales pipeline maturity model

The sales pipeline maturity model

Four levels: Manual, Assisted, Orchestrated, Autonomous. Your level depends on how much of your pipeline keeps running when the person who runs it is away.
Book a discovery call Buy the blueprint
00

The short version

  1. 01

    Your level is your lowest pillar. Score all six, take the worst.

  2. 02

    Level two costs the most. Tools, plus the people moving data between them.

  3. 03

    No purchase fixes this. Each level is a decision somebody wrote down, with a name against it.

  4. 04

    Assume whoever runs outbound was out all last week, then ask what still happened.

01

What is a sales pipeline maturity model?

The Tenbound Sales Pipeline Maturity Model is a diagnostic framework with four levels: Manual, Assisted, Orchestrated, Autonomous. You score it on work that already happened, not on the tools you own.

There is no badge. You get a level, what it costs you, and the next thing to fix.

1 Manual 2 Assisted 3 Orchestrated 4 Autonomous
02

Place your team

Two questions. Question one: take last week, and assume whoever runs outbound was out sick all five days. What happened?

  1. 01 / Manual

    Nothing left the building.

    Outbound depends on one person.

  2. 02 / Assisted

    Sends continued. No new account entered the motion and nothing that came back got answered.

    The tools kept sending. Every decision was still in that person's head.

  3. 03 / Orchestrated

    The queue rebuilt itself, the sequences ran, and the exceptions piled up waiting for them.

    The written rules kept working. Anything that needed judgment waited.

  4. 04 / Autonomous

    The motion adjusted. Plays fired on new signals, losing messages switched off, and a log says what changed and why.

    The system adjusted inside limits a person set.

Question two

Name the document that decided which accounts got worked last week, who owns it, and when it last changed. Most teams answer Assisted to the first question and call themselves Orchestrated. This one settles it.

  1. Manual or Assisted

    There is no document. The account list came out of somebody's head.

    Question one tells you which of the two.

  2. Assisted

    A document exists. Nobody owns it and nothing in it changed last quarter.

    Nobody maintains it, so the tools cannot run on it.

  3. Orchestrated

    A document exists, one named person owns it, and it changed last quarter.

    Someone maintains it, so the tools can run on it.

  4. Autonomous

    The system edited it, and the log says what changed and why.

    The rules maintain themselves inside limits a person set.

Where the two answers disagree, take the lower one. The review starts by checking both, 45 minutes per role.

Two questions against one level scale, highest level first. 04 Autonomous. Question one answer: The motion adjusted. Plays fired on new signals, losing messages switched off, and a log says what changed and why. Question two answer: The system edited it, and the log says what changed and why. 03 Orchestrated. Question one answer: The queue rebuilt itself, the sequences ran, and the exceptions piled up waiting for them. Question two answer: A document exists, one named person owns it, and it changed last quarter. 02 Assisted. Question one answer: Sends continued. No new account entered the motion and nothing that came back got answered. Question two answer: A document exists. Nobody owns it and nothing in it changed last quarter. 01 Manual. Question one answer: Nothing left the building. Question two answer: There is no document. The account list came out of somebody's head. That answer covers Manual or Assisted. Where the two answers disagree, take the lower one.
Level Question one Take last week. Whoever runs outbound was out sick all five days. What happened? Question two Name the document that decided which accounts got worked last week, who owns it, and when it last changed.
04 Autonomous The motion adjusted. Plays fired on new signals, losing messages switched off, and a log says what changed and why. The system edited it, and the log says what changed and why.
03 Orchestrated The queue rebuilt itself, the sequences ran, and the exceptions piled up waiting for them. A document exists, one named person owns it, and it changed last quarter.
02 Assisted Sends continued. No new account entered the motion and nothing that came back got answered. A document exists. Nobody owns it and nothing in it changed last quarter.
01 Manual Nothing left the building. There is no document. The account list came out of somebody's head. Manual or Assisted
The rule

Where the two answers disagree, take the lower one.

Question one says 03Orchestrated
Question two says 02Assisted
Your level 02Assisted
The worked case is the split this section names: question one answered Orchestrated, question two answered Assisted. Assisted is the lower of the two, so Assisted is the level.
03

The four levels

Each level takes one job off a person, always in the same order. Your level is how many of those jobs your team has handed over.

the system holds it a person still does it 1 Draft and send 2 Build and rank the queue 3 Hold the follow up 4 Change what is not working

The four pictures show the same person in the same room. Only the amount of work he does by hand changes.

A man in a beige jacket pinning paper cards to a bright wall by hand, while two white graph8 agents stand idle to his right.
Level 01 / Manual

The list starts blank every Monday.

A rep opens an empty list on Monday and picks who to call. Nobody wrote down which accounts are worth a week, so the rep decides. Another rep would pick a different twenty and nobody would notice.

Human stops
Nothing. A person picks the account, finds the contact, writes the note, remembers the follow up.
System starts
The CRM stores what already happened. A new hire can read all of it and still not know who to call.
How to check
Ask two reps in the same territory for their top twenty accounts. You get two different lists. Neither can be checked against anything.
What this level costs you

Your number is one or two people. When your best rep leaves, their pipeline leaves too. Ramp is slow because there is nothing to hand a new hire. When a month misses there is nothing to open, so the review becomes a conversation about effort.

What to do about it

Write three things in one doc this week. Which accounts count. In what order. What makes a meeting qualified. Buy nothing until that doc exists.

The same man at the same wall, touching one lit card, with a single bone-white agent beside him holding a tablet.
Level 02 / Assisted

AI writes the emails. A rep still picks every account by hand.

Monday still starts with the rep choosing accounts. A model drafts the email and a sequencer sends it. The rep keeps the account criteria in their head and pastes between four tabs.

Human stops
Typing, mostly. The rep still picks every account, approves every send, and remembers every follow up.
System starts
Drafting and enriching in bulk. It runs when someone opens it and stops when they close the tab.
How to check
Open last week's sent folder and ask who read those emails before they went out. If nobody did, this is your level.
What this level costs you

This level costs the most. You pay for the tools and you still pay the people moving data between them. Activity goes up and so does cost per meeting, because the extra volume goes at the same unranked list. Working a bad list faster is how a sending domain gets burned.

What to do about it

Put the account criteria and the qualified-meeting definition in one place the tools read from. Then name who reads AI-written email before it sends.

The same man standing back with his hands down while three bone-white agents work along the wall of cards.
Level 03 / Orchestrated

The queue is built before anyone arrives.

Monday starts with a ranked queue the system built overnight from written rules. The rep works the top of it and escalates anything that does not fit. The rules live in one doc with a named owner.

Human stops
Choosing which accounts to work and in what order. Both decisions now sit in rules somebody wrote and can edit.
System starts
Sequencing. It builds the queue, runs the plays, holds the follow up, and flags the exception instead of dropping it.
How to check
Pull the last ten accounts your team worked. Ask why each was worked that week. Ten answers should point at the same written rule, and anything that does not should already be an exception.
What this level costs you

The playbook stops getting edited. The weekly pipeline review becomes a status read. Conversion sits flat for two quarters while activity looks healthy. The exception queue lands on one ops person, who becomes the bottleneck.

What to do about it

Put two numbers on the weekly pipeline review: meeting to opportunity, and reply quality by segment. Whoever runs that meeting can change the rule that afternoon.

The same man in a long corridor beside a receding line of bone-white agents working a lit wall of cards.
Level 04 / Autonomous

The system changes its own rules inside limits you set.

Nobody opens a list on Monday. A signal fires a play, and the play gets scored against the written definitions. A losing message switches off without anyone booking a meeting about it. A person reads the exception log.

Human stops
Sending, and picking between two versions of a message. The person takes live conversations, exceptions, and the limits themselves.
System starts
Deciding inside those limits. It picks the next action from the definitions it holds, records why, and reports what it changed.
How to check
Ask what the system changed last week. You should get the rule, the result, the date, and who approved the limit it worked inside. If somebody has to reconstruct that, you are not here yet.
What this level costs you

When a definition is wrong, nothing stops. Nobody reads the raw output any more, so it runs for weeks while the dashboard looks fine, and you find it in the forecast a quarter later. The other cost is hiring. Somebody has to write the definitions precisely, and most revenue teams have nobody in that seat.

What to do about it

Give the definitions an owner. They read the exception log every week. They are allowed to switch the system off.

What each level costs

Cost per meeting goes up before it comes down. Level two is the peak. It comes down at level three, where the same send volume goes at a queue somebody ranked, so more of it lands on an account that can buy.

COST PER MEETING higher lower 01 Manual 02 Assisted 03 Orchestrated 04 Autonomous The expensive level

Level two is the long bar. That is the expensive level.

Illustrative. This is the shape of the argument on this page, not a measured series, and there is no scale on either axis. To settle it on your own numbers, take total outbound spend divided by the meetings that became opportunities, one quarter at a time. If that number went up after you bought AI writing, you are at level two.
04

What moves the level

Every level above Manual gets sold as a purchase. Buy the best AI agent on the market while your ICP still lives in three people's heads, and you send more email to the same list. Meetings still come off the account list, and the list did not change. Two decisions set the level, and neither is a purchase.

Two panels on the same level scale, Autonomous at the top down to Manual at the bottom. Left: a purchase arrives, bounces off, and the marker stays at Assisted. Right: two decisions arrive, Who decides which accounts get worked, a data change, and Who fixes it when the number misses, a management change, and the marker moves from Assisted up to Orchestrated.

What gets sold
  • A purchase
  1. 04 Autonomous
  2. 03 Orchestrated
  3. 02 Assisted
  4. 01 Manual

Buying software moves none of these.

What moves it
  • Who decides which accounts get worked a data change
  • Who fixes it when the number misses a management change
  1. 04 Autonomous
  2. 03 Orchestrated
  3. 02 Assisted
  4. 01 Manual

Each level is a decision somebody wrote down, with a name against it.

Illustrative. The marker starts at Assisted because that is where this page says most teams land, and it moves one level because the scoring section says the move is one change at the weakest pillar, run for a quarter. There is no measured series behind either.

Who decides which accounts get worked

  • Manual Each rep, on Monday, from memory. Tier one coverage is whatever they happened to pick.
  • Assisted Still each rep. The AI sends more email at the same unranked list.
  • Orchestrated A written rule RevOps owns. Every rep works the same ranked queue, so coverage is a number you can pull.
  • Autonomous The system, off that same rule. It reranks daily, so nobody works an account that went cold.

Who fixes it when the number misses

  • Manual Nobody, until the QBR. You learn the quarter missed after it ended.
  • Assisted Nobody. Activity is up on the report, so the pipeline review ends without a change.
  • Orchestrated A manager, in the weekly pipeline review. They change the rule that week and read the reply rate the next.
  • Autonomous The system, daily, inside limits one named person owns. A losing message dies before it eats a month of sends.

Moving the first one is a data change: the account rules have to be written down where a system can read them, and that is what the revenue ontology is. Moving the second is a management change: someone has to own the weekly review and change the rule.

05

How to score it

Teams score themselves a level too high because they score the tools they bought, not the work those tools did.

01

Score last quarter.

Score the six pillars on what your team did. Owning a tool does not count. Planning to use it does not count.

02

Your level is your lowest pillar.

Not your average. Strong AI messaging sent to accounts nobody ranked is still a Manual pipeline.

03

Fix one pillar, then score again.

Take the next move for the lowest pillar only. Run it a quarter. Then read the number that pillar owns: connect rate, reply quality, or meeting to opportunity. If it did not move, the fix did not work.

The scoring key

Read across each row and point at the cell that is true for your team. Pillars are the rows, with the number each one moves underneath. Levels are the columns. When two people disagree about the level, use this. The pillars are defined on the Tenbound Standard.

The six pillars of the Tenbound Standard read across the four levels of the Pipeline Maturity Model. Your level is your lowest pillar.
Pillar 01 Manual 02 Assisted 03 Orchestrated 04 Autonomous
Market Coverage of tier one. Reps pick accounts by hunch. A list exists. Nobody can defend it. A written ICP and account tiers set the target. Tiers and territory refresh on live firmographics.
Signal Connect rate. No signals. Every account is a suspect. Intent data is bought, rarely acted on. Triggers rank the queue every morning. A signal fires a play without waiting for a meeting.
Message Reply quality. Every rep writes their own. Templates exist. AI rewrites them unreviewed. Persona messaging carries proof and objection paths. Variants ship, get scored, and retire themselves.
Motion Meeting to opportunity. Sequence by memory. Follow up by mood. Tools send. Nobody owns the motion end to end. One designed motion runs on cadence across channels. Agents run the motion. People take the judgment calls.
Mastery Rep ramp. Onboarding is shadowing. Coaching happens when a number slips. Call and sequence review on a manager cadence. Review covers rep skill and AI output alike.
Measurement Forecast accuracy. Activity counts. Conversion is tracked. Quality is assumed. Meeting and conversion quality measured by source. Continuous, tied to revenue, fed back into the plays.
Read the evidence-based measurement framework Score your team in six questions
06

What Autonomous looks like

Autonomous runs on about ten narrow agents, each with one job and a stated limit, all reading the same definitions. Teams often shop for one system that does everything instead. This is the graph8 operator fleet today.

The Research operator agent: a bone-white humanoid robot with a glowing gradient face screen.
04

Research

Reads the market before anyone acts.

The Signals operator agent: a bone-white humanoid robot with a glowing gradient face screen.
05

Signals

Listens for intent. Surfaces the trigger.

The Enrichment operator agent: a bone-white humanoid robot with a glowing gradient face screen.
06

Enrichment

Raw records in, complete records out.

The Messaging operator agent: a bone-white humanoid robot with a glowing gradient face screen.
07

Messaging

Writes the words that get replies.

The Outbound operator agent: a bone-white humanoid robot with a glowing gradient face screen.
03

Outbound

Moves to the next account. Runs the motion.

The Scheduling operator agent: a bone-white humanoid robot with a glowing gradient face screen.
08

Scheduling

Holds the calendar. Keeps time.

The Lead Ops operator agent: a bone-white humanoid robot with a glowing gradient face screen.
11

Lead Ops

Routes and qualifies every inbound lead.

The Deliverability operator agent: a bone-white humanoid robot with a glowing gradient face screen.
12

Deliverability

Protects the domain. Gets mail to the inbox.

The RevOps operator agent: a bone-white humanoid robot with a glowing gradient face screen.
01

RevOps

Keeps the system honest. Owns the pipeline model.

The Reporting operator agent: a bone-white humanoid robot with a glowing gradient face screen.
09

Reporting

Speaks in numbers. Closes the loop.

Every agent shares one shell, one set of proportions and one material. They are told apart by collar and by a single head accessory, never by colour, which is the same discipline a real fleet needs: the differences have to be the ones that matter.

07

Questions

What is a sales pipeline maturity model?

+

A sales pipeline maturity model describes the levels a pipeline organization moves through as the work stops depending on one person's memory. Tenbound's has four: Manual, Assisted, Orchestrated and Autonomous. Two things set the level: who decides which accounts get worked, and who fixes it when the number misses.

What are the four levels of pipeline maturity?

+

Manual means a rep picks who to call and nobody wrote that decision down. Assisted means tools send at volume while a rep still carries the account criteria between systems by hand. Orchestrated means written rules build the queue and run the sequence, and people take the exceptions. Autonomous means the system picks the next action from definitions it holds, inside limits a person set, and reports what it changed. Buying software moves none of these.

How do I know which level my team is at?

+

Take last week and assume whoever runs outbound was out the whole time. If nothing left the building, you are Manual. If sends continued but no new account entered the motion, you are Assisted. If the queue rebuilt itself and the exceptions piled up, you are Orchestrated. If the motion adjusted and a log says what changed and why, you are Autonomous. If two sound true, name the document that decided which accounts got worked, who owns it, and when it last changed.

Can a team skip a level?

+

No. Each level is built out of what the one below produced. Orchestrated needs the written account criteria that Assisted forces you to admit you never had. Autonomous needs the limits Orchestrated teaches you to set, and the quality numbers that tell a system what winning looks like. Buy autonomous tooling onto a manual foundation and you get faster manual work and a larger bill.

Is Autonomous the goal?

+

Autonomous has its own failure mode. When a definition is wrong, nothing stops it. Nobody reads the raw output any more, so it runs for weeks while the dashboard looks fine. It is only safe when one person owns the definitions, reads the exception log every week, and is allowed to switch the system off.

What does moving up a level require?

+

One change at the weakest pillar, run for a quarter. In most teams the thing holding the level down is a definition nobody wrote down, so the fix is a decision and a document, not a purchase.

08

Where the model came from

Tenbound wrote the model from a decade of sales development campaigns CIENCE ran for these teams, and from the system graph8 runs on itself.

Next step

Talk to an account executive

Thirty minutes with a graph8 account executive, free, or buy the blueprint on the pricing page. Checking your level properly means an engineer watching the work, 45 minutes with one person from each role.

Book a discovery call Score yourself first