Every quarter, Graip.AI moves a little further from “AI that assists” toward “AI that operates.” This recap walks through what’s new and shows how these capabilities are already playing out in real workflows, from contract review to quote reconciliation. Here’s what changed and why it matters.
New Features and Improvements
Hands-Free Agent Interaction
You can control agents using your voice instead of typing, making it faster to send requests when you’re away from a keyboard or juggling multiple tasks. This removes friction for anyone who needs to work with agents on the go.
Email Templates for Agents
You can set up and manage agent email templates directly in the UI, without any manual editing. This standardizes how agents communicate on your behalf, so every outgoing message stays consistent in format and tone across your team.
Usage Monitoring & Management
Account settings now show chat request usage and pages processed, giving teams a clear view of consumption as it happens. This makes it easier to track activity against your subscription limits and plan ahead, rather than finding out after the fact.
Working with Line Items in Tables Directly
When the Quote Management Agent returns a quotation for review, you can edit line items directly in the table, ask the agent to make specific changes, or export the table to Excel, make bulk edits, and upload it back with everything applied. It’s built for the pace and precision RFQ workflows require.
Agent Thinking UI/UX
We have redesigned the agent thinking experience in Agents Lab to give clearer visibility into what an agent is doing at each step – what it’s processing and where it stands in the workflow. This makes it easier to follow an agent’s progress in real time rather than waiting for a result with no insight into how it got there.
Google Drive Storage
Agent-processed files are now stored in shared Google Drive folders, making output versions accessible for review and audits. Teams can trace back through past outputs without relying on the agent’s history.
Open-Source MCP Server
Graip.AI has released an open-source Model Context Protocol (MCP) server for its document extraction engine. Any AI application that supports MCP can now use Graip.AI’s extraction capabilities directly, without custom integration work. This is a shift from Graip.AI’s existing API integrations with SAP and other enterprise systems, which someone has to build and maintain. MCP is a standardized protocol, so any MCP-compatible AI application can call Graip.AI’s tools directly, without custom integration work.
Personalized User Prompt
You can now add a personalized prompt in Agents Lab that supplements the system prompt without overriding it. This enables each user to tailor agent behavior to their own working style and context, without touching system-level configuration.
Agent Evaluation & Monitoring
Graip.AI Agents now include evaluation and monitoring capabilities that systematically measure and improve the quality of agent outcomes. This results in better outcomes with less human intervention in the future.
Spotlight Use Cases
Advanced Quote Management Agent – the “Unhappy Path”
The Challenge: When a manufacturer rejects or modifies items in a customer’s RFP/RFQ – changing quantities, removing items, or substituting products – reconciling those changes against the original request becomes time-consuming. Sales teams must manually compare files, track down what changed, validate it against CPQ data, and figure out how to communicate the revised terms back to the customer.
The Solution: Quote Management Agent that handles the full RFQ/RFP-to-proposal cycle for cases where the manufacturer’s response doesn’t match the original request. When the user forwards an RFP/RFQ, the agent extracts product and line-item data, drafts a clarification email to the manufacturer, and generates a CSV with product details for feasibility checks.
Once the manufacturer returns corrections, the user forwards the reply back to the agent, which automatically reconciles the original CSV against the returned file and validates it against CPQ data.
The agent then presents a clear summary of all changes and drafts a customer-facing email with a revised proposal, which the user reviews and approves before it goes out.
The Outcome: Change tracking, reconciliation, and communication around modified quotes are now automated end-to-end, with the user retaining full control at each key decision point.
Pricing Automation Agent – Label Production
The Challenge: Pricing in the label production vertical involves layered logic – base costs, family groupings, historical benchmarks – that must be applied consistently across every incoming request. Calculating unit costs manually means pulling data from multiple sources, applying pricing logic by hand, and cross-checking against historical benchmarks, all before a single price can be finalized.
The Solution: Pricing Automation Agent that automates end-to-end pricing for label production, from incoming request through finalized unit costs. The agent parses the incoming request, assigns base product costs, groups items by product family, and looks up historical unit costs for comparison. It then calculates unit costs, giving a clear explanation of the logic behind it and presents the figures side by side with the historical benchmark.
The Outcome: Teams get consistent, explainable pricing that accounts for base costs, family groupings, and historical benchmarks automatically, while retaining the final say on which number to use. The result is faster pricing decisions that are grounded in both current calculation and historical context.
Rebate Agreement Agent
The Challenge: Supplier rebate agreements arrive as PDFs in varying layouts and formats. The rebate team has to read each one, classify it by hand into the right category, match it against the correct rate card, manually fill in the discount column, and re-key every line into Pricefx.
The Solution: Rebate Agreement Agent that automates intake, classification, and enrichment end-to-end, escalating only what needs human judgment. The agent extracts agreement data, classifies it into the correct category with a confidence score and supporting evidence, matches it to the right rate card, and fills the discount column line by line. It validates each record for completeness and consistency – clean agreements post straight through to Pricefx, while anything ambiguous is flagged for the rebate manager to review.
After receiving approvals, the agent creates rebate records in Pricefx and notifies the relevant teams on the same day.
The Outcome: The rebate manager only reviews the exceptions, while clean agreements flow straight through from intake to Pricefx posting with full auditability. This removes the manual classification, lookup, and re-keying work that previously consumed hours per supplier, while reducing the errors that come with doing it by hand.
Closing Thoughts
This quarter, Graip.AI focused on making agents more capable, more transparent, and easier to work with day to day. From voice input and personalized prompts to evaluation and monitoring, deeper CPQ and rebate automation, and an open-source MCP server – these updates reflect a platform that’s growing both in what it can do and in how much visibility and control teams have over it.
Explore the new capabilities in your Graip.AI workspace and don’t hesitate to reach out to our team if you have any questions.