Trusted by
AI innovators

We do the work your software leaves to people

We capture the rules, exceptions and judgement your best people carry and build governed systems that execute the work.

BOOK A CALLBBOOOOKK  AA  CCAALLLL
BOOK A CALLBBOOOOKK  AA  CCAALLLL

01

Connect Layer

Plug into Helpdesk, CRM and Chat

Capture requests across channels.

Unified Inbox signals.

01

Connect Layer

Plug into Helpdesk, CRM and Chat

Capture requests across channels.

Unified Inbox signals.

02

Action Layer

Run workflows end-to-end.

Auto-resolve tickets and update records.

Trigger actions across connected tools.

03

Control Layer

Set policies and approval gates.

Route by priority, team, or risk level.

Full audit trail on every action.

04

Context Layer

Pull from knowledge base and past tickets.

Enrich requests with CRM data.

Structured context for every decision.

CONTROL

ANALYZE

CONNECT

RESOLVE

CONTROL

ANALYZE

CONNECT

RESOLVE

PROCESS

KNOWLEDGE

CAPACITY

PEOPLE

WHY WORK STILL FALLS BACK TO PEOPLE

Your rules are in the software.
The follow-through your people do isn’t.

Every system holds a piece of the job. Nobody holds the thread that connects them, not your tools, and not the agents sitting on top of them.

01 / PROCESS

The real process isn’t in one place.

Records live across helpdesk, CRM, inboxes, and documents. SOPs describe the expected path, not every real condition. Exceptions sit in the gaps between systems. People reconnect the work every single day

01 / PROCESS

The real process isn’t in one place.

Records live across helpdesk, CRM, inboxes, and documents. SOPs describe the expected path, not every real condition. Exceptions sit in the gaps between systems. People reconnect the work every single day

01 / PROCESS

The real process isn’t in one place.

Records live across helpdesk, CRM, inboxes, and documents. SOPs describe the expected path, not every real condition. Exceptions sit in the gaps between systems. People reconnect the work every single day

Scenic view of a lush green hillside meadow with a herd of cows grazing under the warm sunlight.
A digital user interface snippet displaying an email forwarding view of an Uber receipt for 28.50 dollars being sent to receipts@payflow.io.
Modify and update the details button.

02 / KNOWLEDGE

The system can’t apply what your people know.

Gather the right context before anyone acts. Check conditions, thresholds, and precedent. Route requests and chase the missing response. Update every system when the work is complete

02 / KNOWLEDGE

The system can’t apply what your people know.

Gather the right context before anyone acts. Check conditions, thresholds, and precedent. Route requests and chase the missing response. Update every system when the work is complete

02 / KNOWLEDGE

The system can’t apply what your people know.

Gather the right context before anyone acts. Check conditions, thresholds, and precedent. Route requests and chase the missing response. Update every system when the work is complete

Expense approval request for Michael Ross's $520 flight.
Sarah Chen profile, auto-approved transaction of $84.
Mike Johnson, auto-approved under $100, for $42.

03 / CAPACITY

More tools haven’t created more capacity.

Every new tool creates another queue to monitor. Handoffs multiply as the operating stack grows. Expert time disappears into status checking. The highest-value judgment work gets squeezed out

03 / CAPACITY

More tools haven’t created more capacity.

Every new tool creates another queue to monitor. Handoffs multiply as the operating stack grows. Expert time disappears into status checking. The highest-value judgment work gets squeezed out

03 / CAPACITY

More tools haven’t created more capacity.

Every new tool creates another queue to monitor. Handoffs multiply as the operating stack grows. Expert time disappears into status checking. The highest-value judgment work gets squeezed out

Watercolor painting of a rolling green landscape with purple heather, cows, and hills.
An interface widget showing automated cloud data integration sync status with QuickBooks Online and Xero logos.
Highlighted Steem app icon in an iOS app dock.

04 / PEOPLE

Your people are still the operating layer.

Capture the logic your best operators already use. Apply approved judgment across connected systems. Route true exceptions to the right person. Keep a complete audit trail for every decision

04 / PEOPLE

Your people are still the operating layer.

Capture the logic your best operators already use. Apply approved judgment across connected systems. Route true exceptions to the right person. Keep a complete audit trail for every decision

04 / PEOPLE

Your people are still the operating layer.

Capture the logic your best operators already use. Apply approved judgment across connected systems. Route true exceptions to the right person. Keep a complete audit trail for every decision

Watercolor painting of a rolling green landscape with purple heather, cows, and hills.
An interface widget showing automated cloud data integration sync status with QuickBooks Online and Xero logos.
Highlighted Steem app icon in an iOS app dock.

WHY ADDING AI HASN’T FIXED IT

AI knows the tool it lives in, not the business around it.

An agent can follow a rule inside its own system. The cross-system context required to act still lives with your people.

01/ FRAGMENTED VIEW

Each agent sees one piece.

The CRM agent sees the opportunity. The ERP agent sees the transaction. The support agent sees the ticket. The process your people run crosses all three.

02 / KNOW-HOW

More data is not more know-how.

A thousand extra records will not teach an agent what to check, who has to sign, in what order, or what proof it needs before it moves.

03 / HANDOFF

The AI stops just short of the work.

It summarizes, drafts, flags, and recommends. Then it hands the job back to the person already doing it to carry the action across the finish line.

Each agent sees one piece.

The CRM agent sees the opportunity. The ERP agent sees the transaction. The support agent sees the ticket. The process your people run crosses all three.

01/ FRAGMENTED VIEW

More data is not more know-how.

A thousand extra records will not teach an agent what to check, who has to sign, in what order, or what proof it needs before it moves.

02 / KNOW-HOW

The AI stops just short of the work.

It summarizes, drafts, flags, and recommends. Then it hands the job back to the person already doing it to carry the action across the finish line.

03 / HANDOFF

Each agent sees one piece.

The CRM agent sees the opportunity. The ERP agent sees the transaction. The support agent sees the ticket. The process your people run crosses all three.

01/ FRAGMENTED VIEW

More data is not more know-how.

A thousand extra records will not teach an agent what to check, who has to sign, in what order, or what proof it needs before it moves.

02 / KNOW-HOW

The AI stops just short of the work.

It summarizes, drafts, flags, and recommends. Then it hands the job back to the person already doing it to carry the action across the finish line.

03 / HANDOFF

WHAT WE ACTUALLY DO

Your systems keep the record. We do the work.

23ML builds a governed system of agents around your operating logic, working across every system the job touches.

It captures the rules, exceptions, approvals and know-how your people carry, then applies that knowledge to run the work end to end.

Workflow example

01

GATHER THE CONTEXT

02

CHECK THE RULE

03

ROUTE FOR APPROVAL

04

UPDATE EVERY SYSTEM

KNOWLEDGE CAPTURE

We encode what your best operators know.

CROSS-SYSTEM EXECUTION

Work moves across your stack, not inside one tool.

APPROVAL GATE

Consequential decisions route to a person.

AUDIT + INSIGHTS

Every action logged with the evidence behind it.

01

Capabilities

CAPABILITIES

The principles behind how the work gets done.

Our platform turns the judgment your best operators carry into a governed system that can move work across every tool, without taking control away from your team.

01 / KNOW-HOW

We build in what your best people know.

Your best operators know which exceptions matter, what evidence an approver needs, and what has to happen next. We turn that into logic the system can apply.

01 / KNOW-HOW

We build in what your best people know.

Your best operators know which exceptions matter, what evidence an approver needs, and what has to happen next. We turn that into logic the system can apply.

02 / CROSS-SYSTEM

The work moves across your stack, not inside one tool.

The agents gather the context, check it against the other systems, take the action, and update the record. Nobody logs into anything new. Nothing is left for your team to stitch together.

02 / CROSS-SYSTEM

The work moves across your stack, not inside one tool.

The agents gather the context, check it against the other systems, take the action, and update the record. Nobody logs into anything new. Nothing is left for your team to stitch together.

03 / JUDGMENT

People step in for judgment, not coordination.

Routine work runs straight through. When a call needs experience, it reaches the right person with the case already built. Context attached. Evidence attached.

03 / JUDGMENT

People step in for judgment, not coordination.

Routine work runs straight through. When a call needs experience, it reaches the right person with the case already built. Context attached. Evidence attached.

04 / OWNERSHIP

We deploy and optimise. You own it.

We build and deploy the layer inside your environment, then keep improving it as the operation changes. Every rule is visible. Every action is logged. The systems, logic, and control remain yours.

04 / OWNERSHIP

We deploy and optimise. You own it.

We build and deploy the layer inside your environment, then keep improving it as the operation changes. Every rule is visible. Every action is logged. The systems, logic, and control remain yours.

WHERE WE START

Three systems that power every workflow

We start in one department, prove it there, then extend the same system into the next. Same chassis, different logic.

We start in one department, prove it there, then extend the same system into the next. Same chassis, different logic.

Revenue

001

Opportunity to cash: deal desk approvals, CRM accuracy, forecast prep, security questionnaires, closed-won handoff.

Finance

002

From transaction to close. Reconciliations, approval chasing, exception resolution, variance follow-up, and month-end prep.

Operations

003

From request to resolution. Ticket handling, vendor and partner comms, policy checks, cross-team handoffs, and service recovery.

Start with the operation that creates the most friction. Prove it there, then extend.
BOOK A CALLBBOOOOKK  AA  CCAALLLL
BOOK A CALLBBOOOOKK  AA  CCAALLLL
BOOK A CALLBBOOOOKK  AA  CCAALLLL

NUMBERS

What changes once the layer runs

Every figure is baselined during the audit and tracked from go-live. No estimates after week one.

DEAL DESK CYCLE

5 days → 4 hours

4 hours

SU

MO

TU

WE

TH

FR

SA

REP ADMIN TIME

Baseline

100 pts

After go-live

76 pts

Net reduction

↓ 24 pts

FORECAST ACCURACY

52% → 78%

139%

INTERNAL IDLE TIME

Before

23 days

After

6 days

CRM DATA ACCURACY

62% → 95%

+95%

FEB

MAR

SYSTEM ACCURACY

87% week one → 98% week twelve

98%

W1

W6

W12

AUTOMATION COVERAGE

With policies + approvals

74%

INTEGRATIONS

Connect AI to the tools your team already uses

We integrate directly with your core systems, so the layer can retrieve data, trigger workflows, and act in real time. Nothing migrates.

Docs & Knowledge

Communication

Service & Support

CRM & Revenue

Finance & ERP

Confluence

Team knowledge and operating documentation

SharePoint

Policies, files, and controlled documents

Google Drive

Shared files and document ingestion

Notion

Workspace knowledge and process context

Docs & Knowledge

Communication

Service & Support

CRM & Revenue

Finance & ERP

Confluence

Team knowledge and operating documentation

SharePoint

Policies, files, and controlled documents

Google Drive

Shared files and document ingestion

Notion

Workspace knowledge and process context

Docs & Knowledge

Communication

Service & Support

CRM & Revenue

Finance & ERP

Confluence

Team knowledge and operating documentation

SharePoint

Policies, files, and controlled documents

Google Drive

Shared files and document ingestion

Notion

Workspace knowledge and process context

SECURITY AND OWNERSHIP

The work runs on its own. Your controls do not move.

GOVERNED BY DEFAULT

Nothing here asks you to loosen a permission, move data somewhere new, or trust a decision you cannot trace.

SOC 2 TYPE II STACK

01/ PERMISSIONS

It works inside the permissions you already grant.

It reaches only the systems, records, and actions allowed for that workflow. If a person could not do it, it cannot do it.

01/ PERMISSIONS

It works inside the permissions you already grant.

It reaches only the systems, records, and actions allowed for that workflow. If a person could not do it, it cannot do it.

02 / GOVERNANCE

Consequential actions stay governed.

Routine work moves automatically. Exceptions, uncertainty, and anything high-impact route to the right person for approval.

02 / GOVERNANCE

Consequential actions stay governed.

Routine work moves automatically. Exceptions, uncertainty, and anything high-impact route to the right person for approval.

03/ TRACEABILITY

Every action is traceable.

What it did, when, on what evidence, and who approved it. Your team gets a record, not an explanation.

03/ TRACEABILITY

Every action is traceable.

What it did, when, on what evidence, and who approved it. Your team gets a record, not an explanation.

04 / DATA

No new system holding your data.

There is nothing new to buy or review. It runs inside the tools you already own, so your data never moves into another vendor’s database.

04 / DATA

No new system holding your data.

There is nothing new to buy or review. It runs inside the tools you already own, so your data never moves into another vendor’s database.

05 / ASSURANCE

Built on an independently audited stack.

Every tool in the system has passed SOC 2 Type II review, ready to withstand enterprise and production rigour at scale.

05 / ASSURANCE

Built on an independently audited stack.

Every tool in the system has passed SOC 2 Type II review, ready to withstand enterprise and production rigour at scale.

06 / OWNERSHIP

You own what we build.

The workflows, the rules, and the logic are yours, deployed in your environment. If the engagement ends tomorrow, you keep all of it. Our IP is in the method, not the output.

06 / OWNERSHIP

You own what we build.

The workflows, the rules, and the logic are yours, deployed in your environment. If the engagement ends tomorrow, you keep all of it. Our IP is in the method, not the output.

CASE STUDIES

Articles, notes, updates, and latest case studies

Insights on support operations, automation design, and the systems behind scalable execution.

Re-building a new-hire ramp at a commercial insurance business

Case Studies

Re-building a new-hire ramp at a commercial insurance business

Case Studies

Illustration of an Object
Re-Building Deal execution for a B2B SaaS + Services

Case Studies

Illustration of an Object
Re-Building Deal execution for a B2B SaaS + Services

Case Studies

Illustration of an Object
Re-building back-office finance across a PE portfolio

Case Studies

Illustration of an Object
Re-building back-office finance across a PE portfolio

Case Studies

Re-building a new-hire ramp at a commercial insurance business

Case Studies

Illustration of an Object
Re-Building Deal execution for a B2B SaaS + Services

Case Studies

Illustration of an Object
Re-building back-office finance across a PE portfolio

Case Studies

FAQ

What leaders want to know before we start

STILL HAVE QUESTIONS?
WE'RE HERE TO HELP

If you handle complex workflows or need custom integrations, our team can help.

CONTACT USCCOONNTTAACCTT  UUSS
CONTACT USCCOONNTTAACCTT  UUSS

Why start with the whole department instead of one workflow?

A workflow rarely operates in isolation. Building one without understanding the surrounding operation moves the bottleneck or creates another disconnected solution. We map the department end to end first, then build each workflow as part of one operating model.

What exactly does 23ML build?

A governed layer around your existing systems. It captures the rules, exceptions, approvals, and operating knowledge your people use, then applies that knowledge to carry work across the business.

How is this different from automation or the AI agents in our existing tools?

Fixed automation follows a predefined path. Application-bound agents understand the one tool they live in. This layer gathers context across systems, applies your company’s operating logic, takes permitted actions, and escalates the decisions that require judgment.

How do you keep the system controlled?

It is built on SOC 2 Type II tooling. Each workflow operates inside defined permissions, approval thresholds, and escalation rules. Routine work moves automatically. Consequential decisions stay governed, and every action is logged for review.

What happens when the system gets something wrong?

Every workflow has a confidence threshold your team sets. Below it, the system does not act. It routes to a person with the evidence and its recommendation. Corrections are absorbed into the logic, so the same mistake does not repeat. Accuracy typically moves from around 87% in week one to 98% by week twelve.

How is this different from hiring a consultancy?

A consultancy hands you recommendations. By week eight we have production workflows removing real work, with logs, approval routing, and a scorecard. The first two weeks look like consulting. Nothing after that does.

What if we want to take it fully in-house later?

You can. The workflows, rules, and logic are deployed in your environment and belong to you. We hand over documentation and train your team. There is no platform to keep paying for.

Why start with the whole department instead of one workflow?

A workflow rarely operates in isolation. Building one without understanding the surrounding operation moves the bottleneck or creates another disconnected solution. We map the department end to end first, then build each workflow as part of one operating model.

What exactly does 23ML build?

A governed layer around your existing systems. It captures the rules, exceptions, approvals, and operating knowledge your people use, then applies that knowledge to carry work across the business.

How is this different from automation or the AI agents in our existing tools?

Fixed automation follows a predefined path. Application-bound agents understand the one tool they live in. This layer gathers context across systems, applies your company’s operating logic, takes permitted actions, and escalates the decisions that require judgment.

How do you keep the system controlled?

It is built on SOC 2 Type II tooling. Each workflow operates inside defined permissions, approval thresholds, and escalation rules. Routine work moves automatically. Consequential decisions stay governed, and every action is logged for review.

What happens when the system gets something wrong?

Every workflow has a confidence threshold your team sets. Below it, the system does not act. It routes to a person with the evidence and its recommendation. Corrections are absorbed into the logic, so the same mistake does not repeat. Accuracy typically moves from around 87% in week one to 98% by week twelve.

How is this different from hiring a consultancy?

A consultancy hands you recommendations. By week eight we have production workflows removing real work, with logs, approval routing, and a scorecard. The first two weeks look like consulting. Nothing after that does.

What if we want to take it fully in-house later?

You can. The workflows, rules, and logic are deployed in your environment and belong to you. We hand over documentation and train your team. There is no platform to keep paying for.