MossaicBook a demo

AI for work that crosses teams and systems

Why does moving work forward take so much chasing?

Mossaic brings AI, people and your systems into one process,
from the first request to the next action.

Book a demo
Follow the work
↗Order release requestORD-2076

Can we release this order today?

Illustrative case · Awaiting review

Recognise the work between the tools?

The request arrives.
The chasing begins.

The order and quote are in one place. Current credit is in another. Finance needs both before deciding.

People end up connecting the pieces by hand.

Order release requestORD-2076

Can we release this order today?

Order + quote$46,800 orderPO-8629 · QT-581
Credit + Finance$40,000 available creditWho can approve the $6,800 excess?
What happens next?

How Mossaic connects the pieces

What if the work
could carry its context?

AI does the groundwork. Your process brings in the rules,
people and systems needed to take it forward.

  • AI agent
  • Business rule
  • Human
  • System
1

Bring the context together

Give AI the whole picture.

AI compares the purchase order with the quote and gathers current account credit. Each finding stays linked to its source.

PO-8629 and QT-581 agree on $46,800. Available credit is $40,000.
Order release requestORD-2076

Can we release this order today?

AI checks the sourcesPrepared
Purchase order · PO-8629$46,800 requested
Quote · QT-581Order details and amount agree
Current account credit · SAP$40,000 available
AI agentFindings linked to their sources

Illustrative case · Sample data · Finance review not yet reached

1 / 3Context
2

Make room for human judgment

Know when a person needs to decide.

The rule identifies a $6,800 excess and routes the request to Finance. The controller reviews the evidence and approves an exception for this order.

Illustrative decision: ORD-2076 only. The standard credit limit stays unchanged.
Order release requestORD-2076

Can we release this order today?

Rule routes to FinanceFinance decides
Source-backed review

$46,800 order · $40,000 credit

The rule calculates a $6,800 excess.
Finance approval required
Finance controller · Sample decisionOne-order exception approved
Recorded

ORD-2076 only. The standard credit limit stays unchanged.

Illustrative case · Sample data · Approval precedes release

2 / 3Decision
3

Carry the decision into the work

Keep the next action connected.

After the approval is recorded, SAP rechecks the order and credit before confirming release. HubSpot confirms the decision note, and the evidence stays with the record.

Illustrative outcome: released for fulfilment, with both system responses confirmed.
Order release requestORD-2076

Can we release this order today?

Systems confirm the actionsConfirmed
Finance approval recorded

ORD-2076 only · $6,800 exception

SAP release confirmed

Order and credit rechecked · REL-2076

HubSpot note confirmed

NOTE-2076 · Evidence linked to CASE-2076

Released for fulfilment. Evidence retained.

Illustrative case · Sample data · Confirmed outcome shown

3 / 3Follow Through
One connected process01 / 03
Order release requestORD-2076

Can we release this order today?

AI checks the sourcesPrepared
Purchase order · PO-8629$46,800 requested
Quote · QT-581Order details and amount agree
Current account credit · SAP$40,000 available
AI agentFindings linked to their sources

Illustrative case · Sample data · Finance review not yet reached

Inside Mossaic

See the decision
with its context.

See the order evidence, Finance’s recorded decision and the confirmed result. The process, decision and history stay connected.

Inside MossaicIllustrated product preview · Sample data · No live actions.

Processes Order release

Order release request

ORD-2076
Released for fulfilment

Process

Process canvasSample version 1
Successful sample · All confirmations recorded

Decision

Order release requestORD-2076

Can we release this order today?

The order exceeds current available credit by $6,800.

Recorded decision · Finance controller
Order amount
$46,800PO-8629 · QT-581
Available credit
$40,000Current account credit from SAP
AI evidence checkSource references retained

Purchase order and quote agree.

AI checked the source documents and retrieved account facts. The rule calculated the $6,800 excess and required Finance review.

Source checksFinance approvalConfirmed release

Recorded authority: ORD-2076 only, for the $6,800 exception. The standard credit limit stays unchanged.

History

Workflow run · ORD-2076

See the decision and its result.

Source checks, the rule result and Finance’s approval stay linked to the confirmed system responses.

Released for fulfilment

Illustrative case. Release, CRM note and evidence retention are confirmed in this sample.

  1. Order and quote checked10:06 · AI agent

    PO-8629 and QT-581 agree on $46,800. Source references retained.

  2. Credit exception identified10:07 · Business rule

    $46,800 order; $40,000 available credit. Finance reviews the $6,800 exception.

  3. One-order exception approved10:12 · Finance controller

    Finance controller approved ORD-2076 only. Standard credit limit unchanged.

  4. Order release confirmed10:13 · ERP

    SAP rechecked the order and credit data and confirmed release. Receipt REL-2076.

  5. Decision recorded in CRM10:14 · CRM

    HubSpot confirmed the decision note. Receipt NOTE-2076.

  6. Evidence record retained10:15 · Run history

    Documents, rule result, approval and both responses linked to CASE-2076.

Illustrative case · All required confirmations recordedSample data · ORD-2076

Key features

A place for the whole process.

Build the route, give AI its part and keep the work visible as it runs.

01

Shape the process

Give every step a place in the plan.

Bring AI tasks, business rules, human reviews and system actions into the same process. Make the dependencies visible before the work runs.

A process canvas, with the next step and its owner in view.Explore in the product tour
ProcessesOrder releaseIllustrated preview
Order release request ORD-2076

Can we release this order today?

Process design
Order review and releaseBPMN-style illustration · Draft design

Scroll across to follow every route.

Order review and release with evidence and Finance holdsStart with the order. An AI task compares the purchase order and quote. Missing or conflicting evidence waits for the case owner to correct it and loops back to the comparison. Complete evidence reaches the credit rule. Within available credit follows configured routine authority. An exception goes to Finance, which can request more information, decline and keep the order held, or approve this order. Both authorized routes enter a collapsed release and follow-up subprocess. It revalidates the order and credit with the conditional SAP release action; a material change requires new Finance review. It records the SAP acknowledgement, then the separate HubSpot acknowledgement and retained evidence. Only all confirmed results reach the recorded final outcome, Released for fulfilment. Any failure or timeout leaves follow-up open. This is a process design illustration.AI evidence taskCompare PO + quoteBusiness rule taskApply credit ruleHuman taskFinance reviewWait for case ownerCorrect evidenceCollapsed subprocessRelease + follow-upRecord decisionKeep order heldRecord final outcomeReleased for fulfilmentEvidence complete?Within credit?Decision?Scope result?Order receivedHeld · DeclinedOutcome recordedFollow-up open

Process design · ORD-2076 follows the Finance exception route. No process is running here.

Subprocess: Release + follow-upDetails
  1. Revalidate and conditionally release in SAP. Enforce the order amount, current credit and recorded one-order authority with the release action. A prior read alone cannot guard concurrent changes. Material changes return to Finance.
  2. Record the ERP acknowledgement. A release is confirmed only with the SAP response. For a failure or timeout, reconcile the order state before retrying with the same case key.
  3. Confirm the CRM note separately. Record the HubSpot acknowledgement. A missing response leaves follow-up open while any confirmed SAP release remains recorded.
  4. Retain evidence and record the outcome. Link PO-8629, QT-581, credit facts, rule result, approval and both responses. Evidence-retention failure also leaves follow-up open.

Recovery preserves confirmed actions and never releases the order twice. “Released for fulfilment” requires every confirmation and the retained evidence record.

Sample dataORD-2076 · Illustrative design
02

Put AI to work

Give the agent context and a clear job.

Choose the information and connector actions assigned to an agent. Keep its contribution connected to the decisions and actions around it.

A defined task, structured inputs and configured connector actions.Explore in the product tour
AgentsOrder releaseIllustrated preview
Order release request ORD-2076

Can we release this order today?

Agent configuration
AI agentOrder review assistant
Illustration

OpenAI · GPT-5

Model for this task
Other provider options
AnthropicOllamaPrivate model

Models depend on your organization’s provider settings.

Read order and account factsSAP · Read actionsAssigned
Approve or release the orderApproval and write actionsNot assigned

Case input: ORD-2076 · PO-8629 · QT-581. SAP account lookup is read-only.

You are the order review assistant.

  1. Compare the supplied purchase order and quote: lines, totals and terms.
  2. Retrieve current available credit with its source reference.
  3. Return matched facts, source IDs and any missing or conflicting evidence.

Flag uncertainty for review. Finance owns exceptions; the process owns release.

Document comparison
PO-8629 and QT-581 agree on $46,800.
Account fact
$40,000 available · SAP account snapshot.
Evidence status
Complete in this sample. Conflicts return for correction.

Sources attached for the rulePO-8629 · QT-581 · SAP account record. The agent cannot approve, change credit or release an order.

Sample dataORD-2076 · Draft configuration
03

See what is happening

See every decision and its result.

Trace the source checks, rule result and Finance approval through to the confirmed ERP release and CRM note. Keep the evidence linked to the run.

A record of who decided, what changed and which systems confirmed it.Explore in the product tour
WorkflowsOrder releaseIllustrated preview
Order release request ORD-2076

Can we release this order today?

Completed
Completed sample6 completed
  1. 10:06 · AI agentOrder and quote checked

    PO-8629 and QT-581 agree on $46,800. Source references retained.

    Confirmed
  2. 10:07 · Business ruleCredit exception identified

    $46,800 order; $40,000 available credit. Finance reviews the $6,800 exception.

    Confirmed
  3. 10:12 · Finance controllerOne-order exception approved

    Finance controller approved ORD-2076 only. Standard credit limit unchanged.

    Confirmed
  4. 10:13 · ERPOrder release confirmed

    SAP rechecked the order and credit data and confirmed release. Receipt REL-2076.

    Confirmed
  5. 10:14 · CRMDecision recorded in CRM

    HubSpot confirmed the decision note. Receipt NOTE-2076.

    Confirmed
  6. 10:15 · Run historyEvidence record retained

    Documents, rule result, approval and both responses linked to CASE-2076.

    Confirmed
Released for fulfilmentSAP confirmed release. HubSpot confirmed the note. All evidence is linked to CASE-2076.
Sample dataORD-2076 · Released for fulfilment
ProcessesOrder releaseIllustrated preview
Order release request ORD-2076

Can we release this order today?

Process design
Order review and releaseBPMN-style illustration · Draft design

Scroll across to follow every route.

Order review and release with evidence and Finance holdsStart with the order. An AI task compares the purchase order and quote. Missing or conflicting evidence waits for the case owner to correct it and loops back to the comparison. Complete evidence reaches the credit rule. Within available credit follows configured routine authority. An exception goes to Finance, which can request more information, decline and keep the order held, or approve this order. Both authorized routes enter a collapsed release and follow-up subprocess. It revalidates the order and credit with the conditional SAP release action; a material change requires new Finance review. It records the SAP acknowledgement, then the separate HubSpot acknowledgement and retained evidence. Only all confirmed results reach the recorded final outcome, Released for fulfilment. Any failure or timeout leaves follow-up open. This is a process design illustration.AI evidence taskCompare PO + quoteBusiness rule taskApply credit ruleHuman taskFinance reviewWait for case ownerCorrect evidenceCollapsed subprocessRelease + follow-upRecord decisionKeep order heldRecord final outcomeReleased for fulfilmentEvidence complete?Within credit?Decision?Scope result?Order receivedHeld · DeclinedOutcome recordedFollow-up open

Process design · ORD-2076 follows the Finance exception route. No process is running here.

Subprocess: Release + follow-upDetails
  1. Revalidate and conditionally release in SAP. Enforce the order amount, current credit and recorded one-order authority with the release action. A prior read alone cannot guard concurrent changes. Material changes return to Finance.
  2. Record the ERP acknowledgement. A release is confirmed only with the SAP response. For a failure or timeout, reconcile the order state before retrying with the same case key.
  3. Confirm the CRM note separately. Record the HubSpot acknowledgement. A missing response leaves follow-up open while any confirmed SAP release remains recorded.
  4. Retain evidence and record the outcome. Link PO-8629, QT-581, credit facts, rule result, approval and both responses. Evidence-retention failure also leaves follow-up open.

Recovery preserves confirmed actions and never releases the order twice. “Released for fulfilment” requires every confirmation and the retained evidence record.

Sample dataORD-2076 · Illustrative design

Controlled autonomy

Give AI more responsibility
within clear boundaries.

Choose the responsibility each agent receives. Keep its work connected to policy checks, named reviewers and the systems around it.

  1. 01
    AI prepares

    Assist

    Read and prepare. The decision stays with the owner.

  2. 02
    AI recommends

    Recommend

    A recommendation reaches the reviewer before an action.

  3. 03
    AI acts within scope

    Act within limits

    The process owner defines the allowed action and approval route.

Your process sets the authority. Configure the tools, policy checks and approval gates around each agent.

Explore the boundaries

Integrations

Bring your tools into the process.

Bring enterprise records, customer context, messages and files into the same process, using the systems your team already works with.

Enterprise and ERP

Bring finance, purchasing and supply chain records into the work.

  • SAP S/4HANAAPI example

Team communication

Send the update to the people who need it, in a channel or by email.

  • SlackConnector setup
  • GmailAPI example

Customer and service systems

Find the customer context and record the next step in the work.

  • HubSpotConnector setup
  • ZendeskAPI example

Data and files

Use the facts and files a decision needs, then store the agreed result.

  • Google SheetsConnector setup
  • PostgreSQLDatabase setup
  • Amazon S3Connector setup

SAP S/4HANA, Gmail and Zendesk are API examples to assess and build. Every connection needs account access, permissions and testing of the required actions.

Tools through MCP

Give agents the tools for the task.

Model Context Protocol (MCP) gives agents a standard way to use tools and data from other systems.

  • NotionDocuments and knowledge
  • LinearIssues and projects
  • GitHubCode and reviews
  • AtlassianJira and Confluence

These providers publish MCP servers. Check compatibility, authentication and the actions each server exposes, then test the connection for your process.

Another system in mind? We can assess its API, webhooks or a suitable MCP server.

Discuss your systems

Security and control

Keep sensitive changes in the right hands.

Choose who can access the work and what an agent can do. Add approval steps before sensitive updates and keep the result with the run.

  • Access for the right people
  • Human approval where it matters
  • A record of the work
See how control works
Order release request · ORD-2076Can we release this order today?
Before approval
  1. Order review assistant

    Check the order evidence

    Compare PO-8629 with QT-581 and retrieve current account credit.

    Complete
  2. Order release policy

    Route the $6,800 exception

    $46,800 exceeds $40,000 available credit. Finance must review.

    Complete
  3. Finance controller

    Review this order only

    Before approval, ORD-2076 stays held. The standard limit is unchanged.

    Waiting for approval
  4. SAP

    Recheck and release the order

    Requires recorded approval and a fresh order and credit check.

    Not reached
Kept with the runIllustrative case · Source checks and rule result recorded. Finance review pending.

A few useful answers

What would you
like to know?

Explore all questions
What does Mossaic do?

Mossaic brings AI assistance, business rules, human decisions and actions in connected systems into a business process. It helps coordinate the steps from an incoming request or changing situation through the decisions and follow-up work.

Where does AI fit into a process?

An agent can be assigned work such as reading information, assembling context or preparing a proposed next step. Rules and process steps define how that contribution is used. The agent's inputs, tools and responsibilities are configured for the task.

Can Mossaic work with our existing systems?

The existing platform includes connectors for tools such as Slack, HubSpot and Google Sheets, plus APIs, databases, email and file storage. Your systems, account permissions and required actions need to be assessed and tested. A connector listing does not mean an account is already connected.

What will we cover in a demo?

We will discuss a process you want to improve and walk through how information, AI assistance, rules, people and system actions can fit together. We can then identify the connections, decisions and implementation questions that need a closer look.

Start with the work on your mind

Where does your
process get stuck?

Bring a process, a recurring exception or a handoff that takes too much chasing. We’ll explore where Mossaic could help.

Book a demo