Navigating a sea of paperwork, scattered throughout varied platforms, generally is a daunting job, typically resulting in gradual decision-making and missed insights. As organizational data and information multiplies, groups that may’t centralize or floor the proper data rapidly will battle to make choices, innovate, and keep aggressive.
This weblog explores how the brand new Discuss to My Docs (TTMDocs) agent offers an answer to the steep prices of data fragmentation.
The excessive price of data fragmentation
Information fragmentation isn’t just an inconvenience — it’s a hidden price to productiveness, actively robbing your crew of time and perception.
- A survey by Starmind throughout 1,000+ data staff discovered that staff solely faucet into 38% of their out there data/experience as a result of of this fragmentation.
- One other examine by McKinsey & Associates discovered that data staff spend over 1 / 4 of their time looking for the knowledge they want throughout totally different platforms akin to Google Drive, Field, or native methods.
The constraints of present options
Whereas there are a couple of choices in the marketplace designed to ease the method of querying throughout key paperwork and supplies residing in a wide range of locations, many have vital constraints in what they’ll truly ship.
For instance:
- Vendor lock-in can severely hinder the promised expertise. Except you might be strictly utilizing the supported integrations of your vendor of selection, which in most situations is unrealistic, you find yourself with a restricted subset of data repositories you possibly can connect with and work together with.
- Safety and compliance issues add one other layer of complexity. When you’ve got entry to 1 platform or paperwork, chances are you’ll not want entry to a different, and any misstep or missed vulnerability can open up your group to potential threat.
Discuss to My Docs takes a special strategy
DataRobot’s new Discuss to My Docs agent represents a special strategy. We offer the developer instruments and help you must construct AI options that really work in enterprise contexts. Not as a vendor-controlled service, however as a customizable open-source template you possibly can tailor to your wants.
The differentiation isn’t refined. With TTMDocs you get:
- Enterprise safety and compliance in-built from day one
- Multi-source connectivity as an alternative of vendor lock-in
- Zero-trust entry management (Respects Present Permissions)
- Full observability by DataRobot platform integration
- Multi-agent structure that scales with complexity
- Full code entry and customizability as an alternative of black field APIs
- Trendy infrastructure-as-code for repeatable deployments
What makes Discuss to My Docs totally different
Discuss To My Docs is an open-source utility template that offers you the intuitive, acquainted chat-style expertise that fashionable data staff have come to anticipate, coupled with the management and customizability you really need.
This isn’t a SaaS product you subscribe to; however reasonably a developer-friendly template you possibly can deploy, modify, and make your personal.
Multi-source integration and actual safety
TTMDocs connects to Google Drive, Field, and your native filesystems out of the field, with Sharepoint and JIRA integrations coming quickly.
- Protect present controls: We offer out-of-the-box OAuth integration to deal with authentication securely by present credentials. You’re not making a parallel permission construction to handle—when you don’t have permission to see a doc in Google Drive, you gained’t see it in TTMDocs both.
- Meet information the place it lives: Not like vendor-locked options, you’re not pressured emigrate your doc ecosystem. You may seamlessly leverage information saved in structured and unstructured connectors like Google Drive, Field, Confluence, Sharepoint out there on the DataRobot platform or add your information domestically.
Multi-agent structure that scales
TTMDocs makes use of CrewAI for multi-agent orchestration, so you possibly can have specialised brokers dealing with totally different points of a question.
- Modular & versatile: The modular structure means you can even swap in your most well-liked agentic framework, whether or not that’s LangGraph, LlamaIndex, or some other, if it higher fits your wants.
- Customizable: Need to change how brokers interpret queries? Modify the prompts. Want customized instruments for domain-specific duties? Add them. Have compliance necessities? Construct these guardrails instantly into the code.
- Scalable: As your doc assortment grows and use instances turn into extra complicated, you possibly can add brokers with specialised instruments and prompts reasonably than making an attempt to make one agent do all the things. For instance, one agent may retrieve monetary paperwork, one other deal with technical specs, and a 3rd synthesize cross-functional insights.
Enterprise platform integration
One other key side of Discuss to my Docs is that it integrates along with your present DataRobot infrastructure.
- Guarded RAG & LLM entry: The template features a Guarded RAG LLM Mannequin for managed doc retrieval and LLM Gateway integration for entry to 80+ open and closed-source LLMs.
- Full observability: Each question is logged. Each retrieval is tracked. Each error is captured. This implies you’ve full tracing and observability by the DataRobot platform, permitting you to really troubleshoot when one thing goes unsuitable.
Trendy, modular elements
The template is organized into clear, unbiased items that may be developed and deployed individually or as a part of the total stack:
| Element | Description |
| agent_retrieval_agent | Multi-agent orchestration utilizing CrewAI. Core agent logic and question routing. |
|
core |
Shared Python logic, frequent utilities, and features. |
| frontend_web | React and Vite internet frontend for the person interface. |
| internet | FastAPI backend. Manages API endpoints, authentication, and communication. |
| infra | Pulumi infrastructure-as-code for provisioning cloud sources. |
The facility of specialization: Discuss to My Docs use instances
The sample is productionized specialised brokers, working collectively throughout your present doc sources, with safety and observability in-built.
Listed here are a couple of examples of how that is utilized within the enterprise:
- M&A due diligence: Cross-reference monetary statements (Field), authorized contracts (Google Drive), and technical documentation (native information). The permission construction ensures solely the deal crew sees delicate supplies.
- Medical trial documentation: Confirm trial protocols align with regulatory tips throughout a whole lot of paperwork, flagging inconsistencies earlier than submission.
- Authorized discovery: Search throughout years of emails, contracts, and memos scattered throughout platforms, figuring out related and privileged supplies whereas respecting strict entry controls.
- Product launch readiness: Confirm advertising and marketing supplies, regulatory approvals, and provide chain documentation are aligned throughout areas and backed by certifications.
- Insurance coverage claims investigation: Pull coverage paperwork, adjuster notes, and third-party assessments to cross-reference protection phrases and flag potential fraud indicators.
- Analysis grant compliance: Cross-reference funds paperwork, buy orders, and grant agreements to flag potential compliance points earlier than audits.
Use case: Medical trial documentation
The problem
A biotech firm getting ready an FDA submission is drowning in documentation unfold throughout a number of methods: FDA steerage in Google Drive, trial protocols in SharePoint, lab reviews in Field, and high quality procedures domestically. The core drawback is making certain consistency throughout all paperwork (protocols, security, high quality) earlier than a submission or inspection, which calls for a fast, unified view.
How TTMDocs helps
The corporate deploys a custom-made healthcare regulatory agent, a unified system that may reply complicated compliance questions throughout all doc sources.
Regulatory agent:
Identifies relevant FDA submission necessities for the precise drug candidate.
Medical overview agent:
Opinions trial protocols in opposition to trade requirements for affected person security and analysis ethics.
Security compliance agent:
Checks that security monitoring and antagonistic occasion reporting procedures meet FDA timelines.
The outcome
A regulatory crew member asks: “What do we’d like for our submission, and are our security monitoring procedures as much as normal?”
As an alternative of spending days gathering paperwork and cross-referencing necessities, they get a structured response inside minutes. The system identifies their submission pathway, flags three high-priority gaps of their security procedures, notes two points with their high quality documentation, and offers a prioritized motion plan with particular timelines.
The place to look: The code that makes it occur
The easiest way to know TTMDocs is to have a look at the precise code. The repository is totally open supply and out there on Github.
Listed here are the important thing locations to start out exploring:
- Agent structure (agent_retrieval_agent/custom_model/agent.py): See how CrewAI coordinates totally different brokers, how prompts are structured, and the place you possibly can inject customized conduct.
- Device integration (agent_retrieval_agent/custom_model/instrument.py): Reveals how brokers work together with exterior methods. That is the place you’d add customized instruments for querying an inner API or processing domain-specific file codecs.
- OAuth and safety (internet/app/auth/oauth.py): See precisely how authentication works with Google Drive and Field and the way your person permissions are preserved all through the system.
- Internet backend (internet/app/): The FastAPI utility that ties all the things collectively. You’ll see how the frontend communicates with brokers, and the way conversations are managed.
The way forward for enterprise AI is open
Enterprise AI is at an inflection level. The hole between what end-user AI instruments can do and what enterprises really need is rising. Your organization is realizing that “ok” client AI merchandise create extra issues than they clear up whenever you can not compromise on enterprise necessities like safety, compliance, and integration.
The long run isn’t about selecting between comfort and management. It’s about having each. Discuss to my Docs places each the ability and the flexibleness into your fingers, delivering outcomes you possibly can belief.
The code is yours. The chances are countless.
Expertise the distinction. Begin constructing at the moment.
With DataRobot utility templates, you’re by no means locked into inflexible black-box methods. Achieve a versatile basis that permits you to adapt, experiment, and innovate in your phrases. Whether or not refining present workflows or creating new AI-powered functions, DataRobot offers you the readability and confidence to maneuver ahead.
Begin exploring what’s potential with a free 14-day trial.
