MaterialWorkflows · Proven patterns across your systems

Let anyone shape the interface. We take care of the seams.

Your people increasingly prompt their own task UIs to get work done. MaterialWorkflows gives that work a trusted way across your systems of record: a governed pattern, and a library of proven use cases drawn from years of enterprise agentic work. Assurance is built in, so it rarely needs to be seen.

Interfaces that come and goA pattern that enduresOnly the seams

Why we begin at the seam.

29%of 897 apps

The average enterprise runs 897 applications. Only 29% of them are integrated.1

58%

Share of the workday knowledge workers say goes to "work about work": chasing, coordinating, updating.2

40%+

Agentic AI projects Gartner expects to be cancelled by the end of 2027, citing cost, unclear value and weak risk controls.3

01 / Our thesis on task UIs

Software that shows up for the task, then lets go.

We think the next interface isn't an app. It's a task UI: an ephemeral, generated screen that anyone can create in the moment, with a prompt, to get one job done. Release these orders. Reconcile this dispute. Onboard this hire. When the job is done, the UI goes away. Nothing to install, nothing to maintain.

01

Made in the moment

The person doing the work describes the job, in the agent they already use. The UI fits that task, not a vendor's roadmap.

02

Calm by design

It sits in the background until judgment is needed, then comes forward with exactly what to decide. Stable shell, generated interior.

03

Let go after

No new app to own. The UI is disposable; what persists is the record in your systems and the audit trail of what happened.

Where this comes from. Sidhant Gupta, co-founder of Material AI, holds a PhD in Ubiquitous Computing and disruptive technologies. We read generative AI through that lens: the goal was always computing that recedes into the work.

"The most profound technologies are those that disappear. They weave themselves into the fabric of everyday life until they are indistinguishable from it."

Mark Weiser · The Computer for the 21st Century · Scientific American, 19914

"If computers are everywhere they better stay out of the way, and that means designing them so that the people being shared by the computers remain serene and in control."

Mark Weiser & John Seely Brown · The Coming Age of Calm Technology, 19965

"Calm technology engages both the center and the periphery of our attention, and in fact moves back and forth between the two."

Mark Weiser & John Seely Brown, 19965
From our research brief

"The 'app' is not destiny; it is what you ship when attention and maintenance are scarce." When building software is expensive, scale wins. When generation becomes cheap, local fit wins.

After the App · Material AI research, 20267

This is no longer speculative. Peer-reviewed systems already generate interfaces from an evolving task model, synthesize widgets on demand that users preferred for repeat edits, and spin up ephemeral UIs that are discarded after the task.6

02 / At enterprise scale

What serves one person well deserves care at scale.

The most valuable task UIs tend to sit on seams: the handoffs between CRM and ERP, HR and IT, service and engineering. Those are also the places where data from two systems of record meets.

In an enterprise, a few familiar questions follow every request. Whose access applies. Which fields should travel. Who signs off. Where the record lives. Answered once, and well, task UIs become something to welcome rather than contain.

Access

Whose permissions apply?

Work runs with the person's own access in each system, nothing broader.

Data

Which fields should travel?

Sensitive details stay private by default, revealed only where policy allows.

Sign-off

Who confirms the change?

Where a decision matters, a named colleague approves it in a single step.

Record

Where is it written down?

Every action is recorded with who, what and why, ready for audit.

Built from Material Plane. Distilled to the seams.

Material Plane is our set of patterns for deploying agents across an enterprise. MaterialWorkflows carries forward only what the boundary between two systems needs, together with the use cases we have seen work, and right-sizes it for one seam at a time.

There is no new platform to adopt. A governed pattern, the use cases that matter for that seam, and connectors to the systems on either side.

Right-sizedOne seam at a timeWorks with the agents you already use
03 / MaterialWorkflows

Interfaces come and go. The way across stays constant.

Whichever agent someone prefers, their work reaches your systems of record in the same considered way, often starting from a proven package. Most of the time, the assurance simply goes unnoticed.

Task UIs · prompted by your people, in their agent of choiceCome and go
Chat assistantYour agent of choice
Show EMEA orders stuck on credit hold and let me release them
Generated UI
Order 2291 · €48k
Order 2304 · €12k
Order 2317 · €210k
Release 2 orders
Completed · within your region
Office suite copilotYour agent of choice
Make a sheet of Q3 renewals with each customer's bank details
Generated UI
AccountRenewsBankAcct 14Aug 30••••••Acct 15Sep 12••••••Acct 16Sep 28••••••
Bank details kept private
IDE agentYour agent of choice
Open Jira issues for this week's escalated support cases
Generated UI
# generated tool
mw.package("MW-04")
 .escalate(cases)✓ 3 issues drafted, linked
Prefilled from package MW-04
Custom internal appYour agent of choice
Issue an 18% credit on this billing dispute and update the case
Generated UI
18% · AR to confirm
Drafted, ready to send
Sent for a light approval
Maintained by Material
MaterialWorkflowsGoverned pattern
Library of proven use cases
Identity & permissionsProven packagesData protectionApprovalsAudit log
  • Activity record
  • REVIEW internal app · credit memo, AR
  • MW-04 ide agent · 3 issues drafted
  • PRIVATE copilot · bank details
  • DONE chat · 2 orders released
System of record ACRMAccounts, orders, cases
The
seam
System of record BERPBilling, credit, ledgers
Systems of record · unchanged, still the source of truthYour systems
The interface belongs to the person doing the work. Shaped in a chat assistant, an office copilot, an IDE agent or an internal app, and set aside when the task is done.
The pattern underneath endures. Maintained by Material, it carries identity, data protection, approvals and a clear record into every request, whichever interface makes it.
04 / How it works

Three layers, each in its place.

The interface belongs to the person doing the work. The systems of record remain your source of truth. MaterialWorkflows sits between them: a governed pattern for working across systems, and a library of use cases proven in enterprise settings over the last few years.

Layer 1 · Task UI

Any agent, any look

People shape the interface they need in the agent they already use. It serves the task, then steps aside.

Layer 2 · MaterialWorkflows

A proven way across the seam

Cross-system work follows one governed pattern, often starting from a ready package. Identity, data protection, approvals and a clear record come with it, consistently, for every interface.

Layer 3 · Connectors

Native to both systems of record

Official APIs and the integration tooling you already run. No separate database. Each system's own permissions and audit still apply.

System of record ACRM, HCM, ITSMRemains the source of truth
System of record BERP, IdP, billingRemains the source of truth
01

Prompt

Someone describes the job in their agent of choice.

02

Generate

The agent builds a task UI for that job, in that moment.

03

Request

The interface calls MaterialWorkflows, often a proven package, for data or an action.

04

Assure

Permissions, data protection and approvals apply quietly, as each package defines.

05

Execute & record

Reads and writes hit both systems; every decision lands in the audit log.

06

Let go

The interface is set aside. The records and their history remain.

05 / The package library

Proven use cases, ready to call.

Each package distills what we have seen work across years of enterprise agentic engagements: connectors for both systems, a rule set, a workflow and the measures that matter. Any task UI can call them, configured to your environment and your approval practices.

MW-01 · SEAM PACKAGE

Quote-to-Cash

Revenue · Finance
Salesforce⇄NetSuite or SAP S/4HANA
The seam
Closed-won deals get re-keyed into ERP sales orders. Terms, SKUs and tax codes drift between systems, and billing starts late.
Rule set
Price and term tolerances checked against the signed contract; discount limits by role; ERP posting only through an approved order.
Workflow
Drafts the sales order from the opportunity, quote and signed contract. Flags price, term and tax mismatches, routes discount exceptions to deal desk, posts to the ERP and writes order status back to the CRM.
Review
Finance approves orders above policy thresholds.
Scored onClose-to-invoice timeOrder exception rateRework touches
MW-02 · SEAM PACKAGE

Hire-to-Retire

People
Workday⇄ServiceNow
The seam
A hire is approved in Workday, then someone opens tickets by hand for laptop, access, desk and payroll checks. Leavers lag even further behind.
Rule set
Access bundles by role and location; non-standard access needs IT sign-off; compensation fields masked outside HR.
Workflow
Turns Workday hire, move and termination events into one onboarding or offboarding plan in ServiceNow, pre-filled by role and location. Tracks every task to done and tells the manager what is still outstanding.
Review
HR confirms the plan; IT approves non-standard access.
Scored onDay-one readinessOffboarding completion timeTickets opened by hand
MW-03 · SEAM PACKAGE

Identity Lifecycle

Identity
ADP⇄Okta
The seam
Role changes in HR never quite reach access. Movers keep old entitlements, leavers keep accounts, and access reviews live in spreadsheets.
Rule set
Least-privilege mapping from HR role to groups and apps; separation-of-duties checks; leavers deprovisioned on policy.
Workflow
Reads ADP joiner, mover and leaver events, proposes the matching Okta group and app changes, flags separation-of-duties conflicts and assembles an evidence pack for every change.
Review
Manager and app owner approve grants; deprovisioning can run on policy.
Scored onTime to deprovisionOrphaned accountsReview evidence prepared
MW-04 · SEAM PACKAGE

Escalation-to-Engineering

Service
Zendesk or ServiceNow⇄Jira
The seam
Support copies tickets into Jira by hand, duplicates pile up, and customers hear nothing until someone remembers to check.
Rule set
Customer personal data redacted before it reaches engineering; one issue per cluster; customer replies need agent approval.
Workflow
Clusters related tickets, drafts one Jira issue with reproduction steps and affected accounts, links both sides, and drafts a customer update whenever engineering status changes.
Review
Support lead approves the escalation; the agent approves each reply.
Scored onDuplicate escalationsTime to engineering responseCustomer update lag
MW-05 · SEAM PACKAGE

Case-to-Cash

Finance · Service
Salesforce Service Cloud⇄NetSuite, SAP or Zuora
The seam
Billing disputes ping-pong between support and accounts receivable over email. Nobody sees the invoice, the contract and the usage in one place.
Rule set
Credit limits by role; payment data masked; every credit memo approved by AR before it posts.
Workflow
Puts invoice, contract terms and usage beside the case, recommends a resolution within credit policy, then drafts the credit memo in the ERP and the reply in the case.
Review
AR approves every credit; larger amounts escalate to the controller.
Scored onDispute cycle timeCredits outside policyDays sales outstanding
MW-06 · SEAM PACKAGE

Campaign-to-Lead

Revenue
HubSpot or Marketing Cloud⇄Salesforce
The seam
Leads cross from marketing to sales through CSV uploads and routing rules nobody trusts. Duplicates and dead leads clog the queue.
Rule set
Consent and opt-outs respected; routing by territory rules; no bulk export of contact lists.
Workflow
Dedupes and enriches on arrival, matches leads to existing accounts, routes by territory and capacity rules, and gives each rep a short why-now brief.
Review
Reps accept or reject with a reason that flows back to marketing.
Scored onSpeed to first touchDuplicate rateLead acceptance rate
MW-07 · SEAM PACKAGE

Account 360 & Field Service

Revenue · Service
Salesforce⇄ServiceNow
The seam
Account teams walk into renewals blind to the open incidents and field visits that live in another system.
Rule set
Read-only by default; contract values visible by role only; any write-back needs the record owner.
Workflow
Builds a pre-meeting brief that joins pipeline, open incidents, SLA breaches and recent field work, and suggests the commitments the account team can actually keep.
Review
Read-only by default; any write-back needs the record owner's approval.
Scored onBrief prep timeRenewal surprisesEscalations raised in QBRs
MW-08 · SEAM PACKAGE

Procure-to-Pay Exceptions

Finance
Coupa or SAP Ariba⇄SAP S/4HANA or Oracle
The seam
Invoices fail three-way match and land in an AP inbox. Clerks chase buyers and receivers by email to clear each one.
Rule set
Match tolerances owned by finance; supplier bank-detail changes always held for review; release needs AP.
Workflow
Explains each exception (price, quantity or receipt), gathers PO and goods-receipt evidence, drafts the fix or the supplier query, and queues it for release.
Review
AP approves every release; tolerance changes are never automatic.
Scored onException agingTouches per invoiceEarly-pay discounts captured

Pairings shown are common examples, not an exhaustive list; the same pattern applies to equivalents such as SAP, Oracle, Microsoft Dynamics or SuccessFactors. Product names are trademarks of their owners and imply no endorsement or partnership. Outcome metrics are what each package is scored on. Targets are set with you during the assessment, not promised in advance.10

06 / Quiet assurance

Assurance, built in quietly.

Good governance should feel like confidence, not friction. It travels with every request, whichever interface makes it. Every package inherits your identity provider, respects each system's permissions and leaves a clear record your risk, audit and security colleagues can read. We design with the controls in frameworks such as the NIST AI RMF and the EU AI Act's human-oversight and record-keeping articles.8,9

Identity & RBAC

The person's own access

Single sign-on through your identity provider. Work runs with the requesting person's permissions, or a narrowly scoped service role, and never more than each system allows.

Human in the loop

Approvals where they matter

Routine work completes on its own. Decisions above agreed thresholds go to a named colleague, in a single step. People can pause or adjust at any point.

Audit & logging

Every read, draft and write

Inputs, model calls, decisions, approvers and writes, logged with lineage. Retention follows your policy and exports to your SIEM.

Data protection

Sensitive details stay private

Field-level masking for PII, compensation and payment data before anything reaches a model. Prompts and outputs checked against your DLP rules.

Data residency

Region-pinned processing

Models and logs deployed in the regions and clouds you choose, including your own tenancy. Your data is not used to train models.

Measured autonomy

Extended with care

New work begins as drafts for review. It moves to running on its own once the evidence, the sign-off and a way back are in place.

The governance layer At the top, the agent surface where people approve, edit or pause. Beneath it, a governance layer of identity and RBAC, DLP, approvals, pause anytime, audit and lineage, and residency and retention wraps the governed workflow. At the bottom, two systems of record remain the source of truth. AGENT SURFACE People review and approve Approve Edit Pause ASSURANCE · TRAVELS WITH EVERY REQUEST GOVERNED WORKFLOW Draft › Approve › Write policy as code · thresholds Identity & RBAC Data protection Approvals Pause anytime Audit & lineage Residency & retention SYSTEM OF RECORD A Source of truth own permissions · own audit SYSTEM OF RECORD B Source of truth own permissions · own audit

Principles we hold to.

Calm interfacesStable shells, with generated detail only where a decision needs attention.
No new dependencyPackages run on your systems and in your cloud, with connectors you own.
Cared-for packagesEach package has an owner, a test suite and a considered end of life.
Persistence by intentInterfaces are set aside by default. What endures is chosen deliberately.
07 / How we begin

Begin with one seam. Extend with confidence.

There is no platform rollout. We begin with a conversation, find the seam that matters most, bring one governed workflow into production, and extend only once identity and assurance are proven in your environment.

01
Consultation

Conversation & assessment

A considered conversation, then a focused assessment of where work crosses your systems of record and where it would matter most to improve it.

  • A map of the seams that matter
  • The use cases worth pursuing, and those worth leaving
  • A view of identity, data and approvals
02
Fixed fee, outcome-first

A first governed workflow

One seam, one pair of systems, in production and available to the agents your people already use. Identity and assurance first, autonomy in time.

  • Connectors, rule set and workflow, configured
  • Review on every change at launch
  • Measures agreed up front, tracked after go-live
03
Extend

Across further seams

Further packages from the library and the assessment, sharing the same pattern, identity model and rules. Each seam becomes simpler to bring in.

  • Next use cases, chosen together
  • Autonomy extended where it is warranted
  • Value measured beyond the pilot, not just usage
"SaaS stays. What changes is what enterprises buy next: fewer sealed features, more governed fit."Material AI research · After the App, 2026
How we charge

Fixed fee. Outcome-first.

Material works with $1B+ enterprises on fixed-fee engagements tied to an outcome we define together, not hours billed or seats sold. We take on a small number of engagements at a time, so senior people stay close to each one.

08 / Begin

The right starting point is a consequential seam.

Tell us about the workflow or decision you want to change across your systems. We will give you a candid view of whether MaterialWorkflows is the right path.

sales@materialcompany.ai

Senior attention throughoutA candid recommendationA limited number of engagements at a time