Home · Solutions · Other solutions
Solution · Other solutionsThe availability answer arrives while the customer is still on the phone
Stock and incoming deliveries answered in Teams, instantly
Sellers, service agents and planners ask in Teams and get free stock, reservations and the next inbound date from live SAP and warehouse data, with the source and time on every answer.
Executive summary
A ten-second lookup costs two people ten minutes each, and by then the customer has asked somebody else.
We build a conversational agent with UiPath Agent Builder and publish it to Microsoft Teams as a Teams app.
Availability is answered while the customer is still on the call, because a lookup that takes seconds no longer waits for a person to be free.
SAP S/4HANA with Extended Warehouse Management (read-only); the logistics provider's warehouse system (read-only); Microsoft Power BI
Business problem
Inventory
Stock information exists in three places. SAP holds the two own warehouses, the third-party logistics provider's system holds the Romanian one, and a planner's spreadsheet holds the quantities already promised. Sellers and service agents either have no SAP licence or have one and cannot read storage locations, reservations and available-to-promise logic. So the warehouse channel in Teams and the planning mailbox become the interface, staffed by people whose actual job is receiving, picking and planning.
Every question interrupts somebody, every answer is typed by hand, and the answer is out of date by the time it is read. The same part number is asked about by three sellers in one afternoon and answered three times. Nothing is recorded, so the planner never learns which parts sellers are struggling to promise, and the channel never gets better at its job.
The reason this persists is that the obvious fixes do not work. Giving everyone SAP access is expensive and does not solve the readability problem. A Power BI stock report refreshed overnight is honest but too old for a customer conversation. Nobody owns the question channel, so nobody improves it, and it quietly becomes the most important interface in the company that has no service level attached to it.
How it works today
- PersonA customer asks a seller whether a quantity is available and when it could ship
- PersonThe seller messages the warehouse channel or emails planning, then goes back to the call
- WaitingThe coordinator finishes the task in hand before opening SAP, because picking does not pause
- SystemStock per storage location and open reservations are checked, then inbound purchase orders for the next delivery
- WaitingIf the part sits in the third warehouse, the logistics provider is emailed and answers when it answers
- PersonThe answer is typed into the channel, by which time the customer has often moved on
- Risk of errorNothing is logged, so the next seller asks the same question tomorrow and the promise made from memory reaches the warehouse at picking
Why the current process costs more than it appears
The cost grows where nobody is looking.
- Quotes go out with "subject to availability" and lose to whoever answered. That loss never appears in a report, because nobody records the enquiry that arrived too late.
- Promises made from memory are discovered at picking as shortages and split shipments, which cost twice: the expedited freight and the phone call to the customer who was told something else.
- Planners interrupted twenty times a day produce worse forecasts and leave sooner. Recruiting a planner costs more than a year of the channel they are being used as.
- Discrepancies between SAP and the logistics provider's system stay invisible until a customer complains, because nobody compares the two systematically.
- When the two coordinators who answer the channel take leave in the same week, the sales floor slows down and nobody has planned for it.
Cost of inaction
Guessing is what replaces a slow channel, and guessing is free until picking. As enquiries grow with the product range, the customer base and every new sales hire, the number of people who can read SAP stays where it is, so the channel gets slower and sellers quietly stop asking. The warehouse then meets the guesses on the floor, as shortages, split shipments and expedited freight that nobody attributes back to a question that went unanswered.
Adding a fourth warehouse, a marketplace or a second ERP instance makes the consolidation worse rather than better, because each new system adds another person to ask. None of this appears as a line item. It appears as quotes that lost on speed, freight that was not planned, and a planning team that keeps recruiting to stay level.
A plausible organisation with realistic proportions. The figures are there to be recalculated on your data; they are not a client result.
An industrial components distributor in fasteners, hydraulics and transmission parts, 620 employees, two warehouses in Poland on SAP S/4HANA with Extended Warehouse Management and a third in Romania run by a logistics provider on its own system. Microsoft 365 is the collaboration platform.
Around 2,900 stock and delivery enquiries a month from 14 inside sales staff, 22 field representatives and nine customer-service agents, answered by three planners and four warehouse coordinators between their other duties. A nightly Power BI stock report exists and is rarely opened.
Questions go to the warehouse Teams channel or the planning mailbox. Somebody opens SAP, checks storage locations and reservations, emails the logistics provider for the Romanian stock, checks inbound orders and types a reply.
About 11 minutes consumed per enquiry across both sides, waiting and re-asking on one, lookups and email on the other, and roughly a third of enquiries pull a planner out of what they were doing.
A UiPath conversational agent published to Microsoft Teams interprets the question, queries SAP and the logistics provider's system through deterministic tools, and answers with quantities, reservations, the next inbound date, the source and the time. Anything that needs an allocation decision becomes a task for a planner.
In the modelled case the seller answers on the first call, planners are interrupted for decisions rather than for lookups, and the question log becomes a demand signal the ERP does not have. Illustrative, not a client result.
Proposed solution
We build a conversational agent with UiPath Agent Builder and publish it to Microsoft Teams as a Teams app, so the seller asks in the same place they ask a colleague today. The agent's job is deliberately narrow: work out what is being asked, resolve the part number and any alias the sales team uses, identify the location and the question type. It does not know any numbers and is not asked to.
The numbers come from tools. UiPath API workflows query SAP through the OData connector for stock by plant and storage location, open reservations, sales orders and inbound purchase orders, and a connector built with UiPath Integration Service Connector Builder reads the logistics provider's system for the Romanian warehouse. Every answer carries the quantities, the reservation position, the next inbound date, the source system and the time of the query. If the question is ambiguous, the agent asks rather than guesses; if it falls outside stock and delivery, a guardrail declines it.
Decisions stay with people. Reserving stock for one customer ahead of another, or expediting a delivery, becomes a UiPath Action Center task that the planner completes inside Teams, with the enquiry attached. A scheduled robot posts an Adaptive Card to the sales channel when a watched item drops below its reorder point or an inbound date moves, so the seller who promised it hears before the customer does. Every question and answer is written to UiPath Data Fabric, and Power BI turns that log into a planning view: which parts, which locations, which weeks.
UiPath conversational agents published to Microsoft Teams; UiPath Agent Builder tools, contexts and guardrails; UiPath Action Center tasks completed in Microsoft Teams; UiPath Integration Service SAP OData and Microsoft Teams connectors; UiPath Orchestrator triggers, credential store and job audit; UiPath AI Trust Layer
The question taxonomy from your own channel, the agent's prompts, tools and guardrails, the SAP query workflows, the connector to the logistics provider's system, the escalation rules into Action Center, the alert rules for watched items, the question log model and the Power BI planning view
SAP S/4HANA stock, reservation, sales-order and inbound queries through the OData connector under a read-only service user; the logistics provider's warehouse system through a connector built with UiPath Integration Service Connector Builder
How the automated process works
- PersonA seller asks the agent in Teams, in the words they would use with a colleague
- AutomationThe agent resolves the part number, alias and location and classifies the question type
- SystemAPI workflows query SAP for stock, reservations, sales orders and inbound deliveries
- SystemA second tool reads the logistics provider's system when the part sits in the Romanian warehouse
- AutomationThe answer comes back with quantities, reservations, the next inbound date, the source and the timestamp
- PersonReservation and expediting requests become Action Center tasks in Teams for a planner to decide
- AutomationA scheduled robot posts a card to the sales channel when a watched item's stock or delivery date moves
- AutomationEvery question and answer is logged and surfaced in the Power BI planning view
Human-in-the-loop model
Automation handles
- Interpreting the question and resolving part numbers, aliases and locations
- Querying SAP and the logistics provider's system for stock, reservations and inbound deliveries
- Composing the answer with its source and timestamp, and asking for clarification when unsure
- Logging every enquiry and alerting the sales channel on watched items
People decide
- Allocation between customers when free stock will not cover both
- Expediting with suppliers and with the logistics provider
- What to do about the discrepancies the agent surfaces between the two systems
- Which answers were wrong or unclear, reviewed as part of the agent's evaluation set
Before and after
Systems and integrations
We do not add technology to make an architecture look serious. Every element below has a specific job in this process.
Inputs
- questions asked in Microsoft Teams
- the part-alias lists the sales team already uses
- watched-item rules from planning
Automation layer
- UiPath Agents
- UiPath Orchestrator
- UiPath Integration Service
- UiPath Action Center
- UiPath Data Fabric
Target systems
- SAP S/4HANA with Extended Warehouse Management (read-only)
- the logistics provider's warehouse system (read-only)
- Microsoft Power BI
Human touchpoints: the agent conversation in Teams; Action Center tasks in Teams for planners; the alert card in the sales channel
Technologies used
interprets the question and calls the tools; publishes into Teams as an app
Astock by plant and storage location, reservations, sales orders, inbound purchase orders
Athe read-only connector to the logistics provider's warehouse system
Areservation and expediting requests as tracked tasks with an owner
Aallow-listed model, EU region routing, audit of prompts and tool calls
Athe alert job, the credential store and the audited job log
Athe planning view over the question log and the discrepancies between systems
Anumbers come only from query results; out-of-scope questions are declined, not answered
CIllustrative economic model
What it is worth, with the arithmetic shown.
An enquiry, not a person, is what the 11 minutes below belong to: waiting and re-asking on one side, the SAP lookup, the email to the logistics provider and the typed reply on the other. The calculator prices only the share an agent can answer alone, so its volume is 0.8 of 2,900 enquiries rather than all of them, at €34 an hour blended across sales, service and warehouse roles. Planner refocus time sits beside it: 0.35 of the enquiries × 4 minutes is about 68 hours a month at €46, some €37,500 a year, taking the combined pool to roughly €211,000. These are our assumptions for a distributor of this shape, not measurements.
Run the numbers on your data
An illustrative estimate from your own inputs. It models released capacity; it is not a promise of savings.
Business benefits
- Availability is answered while the customer is still on the call, because a lookup that takes seconds no longer waits for a person to be free
- Planners and coordinators get their day back and are interrupted for decisions rather than for readings from a screen
- Over-promising falls, because the seller sees reservations and inbound dates instead of remembering a number from last week
- Differences between SAP and the logistics provider's system surface on the first question rather than at the customer complaint
- Planning gains a demand signal the ERP does not hold: which parts are hard to promise, where and how often
- Holiday and absence stop being a sales problem, because the answering layer does not take leave
The management view
- Questions that used to vanish into chat history become countable: by part, location, asker and week
- Reservation and expediting requests become tracked tasks with an owner and a response time instead of favours asked in a channel
- Sales management can see how many customer conversations ended with an answer on first contact
- The logistics provider's data quality becomes measurable, because every mismatch is recorded when it is found
Board-level KPIs
Security and governance
Where the data sits and who can see it.
- The agent's tools run under a read-only SAP service user and a read-only key for the logistics provider's system, so nothing in the chat can change stock, a reservation or an order
- Those secrets sit in the Orchestrator credential store or in Azure Key Vault, never in a workflow or a prompt
- Model access runs through the UiPath AI Trust Layer with an allow-listed model, EU region routing and an audit record of every prompt and tool call
- Answers are restricted to what the tools returned, guardrails decline questions outside stock and delivery, and answers can be scoped to the asker's sales region
- Users are authenticated by Microsoft Entra ID through Teams; the log keeps the question, the answer and the identity of the asker, and no other personal data
Why now
Conversational agents can now be published into Teams as an app, and SAP OData connectors turn stock, reservation and inbound queries into configuration rather than screen automation; neither was true a few years ago
Buyers compare every supplier with the availability answer a web shop gives them in one second, and sales teams are being asked to respond on first contact
Distributors keep adding third-party sites and marketplaces, so the number of systems holding stock rises each year and the case for a single answering layer rises with it
Relevant executive roles
Planners and coordinators stop working as a help desk, and over-promises stop arriving on the warehouse floor as a surprise
The question log is a demand signal the ERP does not produce, and discrepancies with the logistics provider become measurable
Sellers answer availability on the first call and quote against real reservations and real dates
Common questions and objections
Some can. Most cannot read storage locations and reservations, and the third warehouse is not in SAP at all. The agent reads all of it and answers in the words a seller uses.
The model produces no numbers. It works out what is being asked and calls a query; the answer is the query result with its source and timestamp, and an unclear question comes back as a clarification rather than a guess.
Then the agent makes that visible in the first week, which is where fixing it starts. Reconciliation between the warehouse system and the ERP is a companion piece of work, not a prerequisite for asking better questions.
When this is not the right solution
- One warehouse, one system, and everyone who needs stock information already has a screen they can read
- Stock accuracy is low enough that any answer would mislead; the data work comes first and this comes after
- Enquiry volume is small enough that a nightly stock report genuinely satisfies the sales team
A question for the next management meeting
Our sellers ask the warehouse the same handful of questions nearly three thousand times a month: what is that costing us in quotes answered too late, and who in this company owns that channel?
Implementation approach
What we deliver, and what we need from you to start.
We deliver
- Two to four weeks of your channel and mailbox enquiries, grouped into the question types that actually occur
- The agent: prompts, tools, guardrails, clarification behaviour and the evaluation set we test it against
- The SAP query workflows and the connector to your logistics provider's system
- Escalation rules into Action Center and alert rules for watched items
- The question log model and the Power BI planning view
- A pilot with one sales team on SAP data only, measured against the channel it replaces
We need from you
- Read-only SAP access for a service user, and the logistics provider's API documentation
- A planner as product owner, able to decide what the agent should and should not answer
- The part-alias lists your sellers use, however informal they are
- Two to four weeks of channel messages and planning emails
Stages
Discovery
Real questions from the channel, grouped; usually a handful of types cover most of the volume
Design
Agent scope, tools, guardrails, escalation rules and the answer format
Build
SAP query workflows, the logistics-provider connector, the agent, alerts and the log
Validation
The agent tested against real historical questions, with answers checked by planners
Pilot
One sales team, SAP data only, against the channel it replaces
Extension
The third warehouse, the alerts and the remaining teams
Quick win. Effort is driven by the number of source systems and the state of part-number aliases, not by the agent itself; the third-party warehouse connection is usually the longest item.
Eleven minutes for a ten-second lookup, and the customer did not wait.
Give us two weeks of messages from your warehouse or planning channel. We come back with the share an agent could answer straight from SAP and your warehouse system, and the questions that will always need a planner.
Give us two weeks of stock questionsThe neighbouring process usually has the same problem
Stop discovering at the cycle count, or when an order cannot ship, that the WMS and the ERP disagree.
View solution Customer serviceOrder status answered before the customer asksEvery third customer contact asks where an order is; the answer already sits in SAP and the carrier portal.
View solution Other solutionsFinance questions answered in Teams from governed numbersThree versions of ARR are in circulation, and the board saw one of them last quarter.
View solutionIndustries we deliver this in most oftenManufacturing & industryRetail & e‑commerce