Fall Release ’26: Agentic AI for AP Teams and Sellers

Published on: September 24th, 2026

We’re extending agentic AI across premium reporting, AP auditing, document intelligence, and self-service support, all built on one secured foundation.

 

If Spring was about extraction and reporting, Fall is about turning that data into something your team can act on without opening a ticket. This release adds agentic AI across the board: sellers get sharper payment predictions, more reports and dashboards, and instant answers to “where is my payment?”, while buyers get their own AP Auditor to talk to, finer control over how documents are extracted, and a full audit trail to go with it.

On September 16th we had our customer dedicated Fall Release webinar, and it was, as usual for our release webinars, a highly attended and well-rated session. AI capabilities were the second most requested topic for a follow-up session, just behind e-invoicing compliance. If you registered but missed the live session, or want to watch it again, it’s now available on our on-demand webinar page.

Here’s a recap of what was presented during the webinar and why it matters.

For a full picture of everything in this release, head over to our Release Notes on the Knowledge Base.

 

Reporting & Analytics App for Sellers: Four New Reports

Launched in June 2026, the Reporting & Analytics App for Sellers with a paid subscription now includes four new reports and dashboards: Document Header level, Document Lines level, Users, and Connections. Your invoicing performance, team adoption, and buyer relationships now live in one app.

What this means for your team:

  • Accelerate order to cash: track the invoice lifecycle in real time, spot approval bottlenecks, and forecast incoming cash accurately
  • Optimize inventory and fulfilment: analyze purchasing trends to forecast demand and align stock with buyer needs, for fewer stockouts and less overstock
  • Strengthen buyer relationships: monitor buyer purchasing habits to proactively offer tailored contracts, volume discounts, and personalized support

A Sharper Payment Predictor Model

On September 1, 2026, we launched an improved Payment Predictor Model for sellers, built to handle payment volatility and business seasonality better than the previous version. Sellers now get clearer visibility into when they’ll be paid, better cash flow planning, earlier flags for late or early payments, and fewer surprises.

The model adapts to each seller and buyer relationship rather than applying a single global average, which means more trustworthy confidence scores and higher accuracy on the invoices it does predict. It’s also more inclusive: roughly 1,200 sellers are now eligible to access predictions, up from about 700 under the old model, and more invoices get a prediction overall because the model is more precise about which relationships are predictable.

What this means for your team:

  • Nearly double the eligible sellers: about 1,200 sellers can now access predictions, versus 700 previously
  • Predictions tuned to the relationship: the model learns each seller and buyer pairing instead of applying one blanket average
  • Better planning, fewer surprises: clearer signals on when payment is coming, with earlier flags when it’s running early or late

 

Meet the Tradeshift AP Expert Auditor

Also launched in June 2026 and available in the Premium version of the Reporting & Analytics App, the AP Auditor is a specialized chat agent built once and available continuously. It answers questions, runs complex analyses, and takes actions on the platform through the Tradeshift MCP Server.

It reads the PDFs in its workspace for rules and context, pulls live data, and executes changes through the same permissions the requesting user already has, so the system stays authoritative on what each role can and cannot do.

What this means for your team:

  • A specialist on call: complex analyses and actions, available continuously rather than scheduled around a person’s availability
  • Grounded in your own rules: it reads the documented processes and context you give it, rather than applying generic logic
  • No permission workarounds: every action runs through the same access controls the requesting user already has

AI Document Intelligence: More Control, Full Traceability

We shipped four upgrades to AI Document Intelligence this release, each aimed at giving your team more control over extraction and a clear record of what changed and why.

Custom Prompt Configuration (launched July 2026) lets your PS team tailor AI extraction per customer, and even per individual supplier connection, with no code and no deployment. Buyer level changes add or remove extracted fields and override rules across all of a buyer’s invoices; connection level changes apply supplier specific overrides to one connection only, with zero impact elsewhere. If one supplier labels a field “Ship Date” instead of “Delivery Date,” that gets mapped on their connection alone, without touching anyone else.

Business Audit Trail (also launched July 2026) now records every AI extraction event and every configuration change. Extraction attempts are logged as started, succeeded, or failed, along with the model used and its confidence score. Configuration changes are logged with before and after values, who made the change, and when. It’s built for enterprise audit requirements, not a black box.

Invoice PO Reference Auto-Enrichment (launched July 2026) autofills the PO number and PO line reference on invoice lines the moment the PO is available, removing a manual “view PO” step that used to stall invoices in clearance flows like Poland and France while waiting on a buyer side PO push.

Confidence-Threshold Auto-Dispatch (launched September 2026) lets you set a confidence threshold for AI-extracted fields. Invoices that meet it are dispatched automatically with zero manual touch; documents below the threshold get a clear signal, with the AI status icon switching from purple to yellow and a warning banner explaining what needs attention. Automation rate is tracked live, so you can benchmark your straight-through performance.

What this means for your team:

  • Extraction tuned to your suppliers: buyer and connection level overrides handle exceptions without a single line of code
  • Full traceability: every extraction attempt and every configuration change is logged, with who, what, and when attached
  • Fewer manual touches: PO references auto-fill and high-confidence invoices auto-dispatch, so your team only sees what actually needs review

 

AskAda for Sellers: “Where Is My Payment?”

Integrated directly into the Tradeshift platform, AskAda for Sellers lets suppliers self-serve. They can query real-time platform data in plain language, in their own language, and get accurate answers in seconds, with no support ticket and no waiting.

What this means for your team:

  • Data availability: suppliers get the current status of any invoice, payment, or PO without contacting your team
  • Faster resolution: answers arrive in seconds, in the supplier’s own language
  • A better seller experience: instant, real-time updates reduce delays in resolving invoice-related issues

 

The Bigger Picture: Procure-to-Pay Financial Intelligence

Everything in this release sits on the same foundation: AI Document Intelligence, Agentic Analytics, Compliance, and an open Agentic Platform, four pillars turning every document on the network into something a finance team can act on.

AI Document Intelligence handles any supplier, any layout, any PDF, with every line coded at scale and no templates or onboarding required. Agentic Analytics takes natural language questions from any persona and returns answers over 1 million to 100 million rows in seconds, no BI project needed. Compliance covers clearance, Peppol, and mandates with business rules set per customer. And the platform itself is an open agentic hub: 95 tools available over open APIs through MCP, so you can point your own agents at your own data, not only ours.

Underneath all of it is governance, efficient operations, quality data access, scalability, and security: one AWS perimeter, agents that inherit the user’s own permissions, and an immutable audit trail.

For a full picture of everything in this release, head over to our Release Notes on the Knowledge Base. And if you’d like to watch the full webinar, it’s available on demand here.

 

Curious how these AI capabilities could work in your environment? Reach out to your Tradeshift Customer Success Manager if you are already a Tradeshift user, or contact sales here to schedule a demonstration.