Our Agentic Team at Workato has learned a lot this year from working with our customers, especially our CIO community.
A key theme in their feedback has been the need to manage the risks that companies associate with making their first investments in Large Language Models and Generative AI. These are the concerns that, left unaddressed, keep early AI efforts from growing out of the POC phase and providing real business value at scale.
The good news is that Workato began building the kinds of tools and controls that organizations can use to manage this AI risk over a decade ago, and are already part of our DNA. Our Agentic Platform, which is coming to customers in the month of June as part of Workato ONE, adds more tools for building and deploying Agents safely in the business.
Three types of AI Risk:
- AI Model Risk – These are the risks inherent in using Large Language Models themselves in a business setting. The size and complexity of an LLM’s ‘latent space’ leads to answers and outputs which are fundamentally less rules-based and predictable than the previous machine learning models that enterprises used for things like Credit, Risk and Fraud. They are trained with gigantic, web-scale amounts of data and are typically partially or completely closed-source, making it difficult to understand how they reach specific conclusions by looking ‘into’ the model itself. That lack of explainability and repeatability can cause the kinds of mistakes in an enterprise environment – especially in regulated data situations – that create legal and compliance issues. No CIO wants to risk lawsuits, regulatory fines, or the company’s use of AI creating a dreaded news headline.
- AI Market Risk – CIOs feel pressure to ‘do something’ about AI right now. This urgency comes from the CEO, the Board, employees, and shareholders – sometimes all at once! The challenge is that companies have to evaluate AI initiatives and investments in a very fast-moving technology market. New model versions and new model capabilities are announced every week. New pricing models and open source alternatives emerge on a regular basis. Even the ‘best practice’ architecture and tech stack on which to build is a moving target as new technologies and use cases emerge. This dynamic environment is exciting but it also creates analysis paralysis for the CIO. If I go all-in on one LLM provider and architecture, will next week’s technology news make my investment look foolish and obsolete?
- AI Execution Risk – The first question we hear from many of the CIOs we meet is where and how to start. Companies feel urgency to get the business impact promised by GenAI and LLMs but there still is not an established playbook for the minimum viable architecture and the best use cases to deploy first in the business.
For any new area of IT investment, an executive has to find the sweet spot of:
- Cost
- Relevance to the business
- Time to value
- Governance (internal feasibility)
This need to balance many different decision criteria and identify the right skill-sets often leads to analysis paralysis. After months of meetings, nothing actually gets done.
At Workato, we are building our platform to help our customers manage all three of these risk areas.
Managing AI Risk With Workato
- Skills – By using Workato’s industry-leading connectivity and orchestration capabilities to build Skills, companies can create simple rules, conditions and error handling for the systems and data an Agent wants to access. Without writing a single line of code, Skills allow open-ended interaction and autonomy for AI Agents without exposing their business systems to risk. This creates a balance between the probablistic and ‘open-ended’ nature of the LLM and the deterministic and repeatable nature of Workato’s Orchestration.
- Agent Controls – Workato’s Agentic Platform lets builders create specific instructions for LLM’s about how to handle certain situations, escalate, and use company policies. These controls together with Workato Skills, make it easier to get safe and predictable outcomes from AI Agents that fit into the company’s compliance strategy.
- Workato ONE – Everything that Workato already does to enable enterprise orchestration applies to the Agentic Platform too. From connectivity and secrets management, to custom RBAC, to logging and observability, Workato allows customers to integrate AI Agents into an overall IT and governance strategy.
Managing AI Market Risk With Workato
- Flexible AI Architecture – Workato is not a monolithic system of record that’s best known as a CRM or an ITSM company. Workato is not a company that’s best known as an AI Research Lab, making Large Language Models. This allows Workato to offer the Agentic Platform to customers in a neutral and flexible way. If a better, cheaper, or open source LLM becomes available then companies can incorporate it into their AI workflows and Agentic applications without missing the market. Our CIOs are protected from making obsolete investments.
- Open Standards – Going forward a company’s AI Agents may come from many different sources – they probably won’t all be built in Workato’s platform. As the AI landscape evolves, new standards like Model Context Protocol (MCP) and Agent To Agent (A2A) are emerging to help companies build and integrate Agents. Workato already supports MCP, and will support A2A this Summer. These standards allow Workato to be the system of record for the flow of Agents across a company’s systems and data.
Managing AI Execution Risk With Workato
- AgentX Apps – Every company has an IT department, an HR department, and other core functions like Sales and Finance! To help our customers deploy AI Agents faster, Workato offers pre-built AgentX Apps for these functions that can be configured and launched to create immediate business value. This makes knowing where to start easy.
- Fully-Managed Knowledge Base – For a company to get business value from AI and LLMs, data and information sources have to be incorporated into an AI-ready Knowledge Base. To prevent this process from blocking AI initiatives, Workato offers an end to end RAG and Knowledge Base. Companies just point to the sources of their information, data and policies – and Workato does the rest.
- Agent Studio – Vibe coding and Agent Frameworks can be a great way to hack AI prototypes together, but these experiments lack the tools, governance and scale needed to make it into production. Workato Agent Studio lets builders assemble the key Skills, Models, and Knowledge needed to create Enterprise AI Agents for new use cases without sacrificing speed and scale.
- Workato ONE – Some use cases are best served by an automated workflow that uses an LLM but still follows a fixed set of steps. Other use cases are a great fit for open-ended Agents that have more autonomy and make more decisions. Workato ONE enables companies to select the best tools for each use case without needing to worry about which vendor and architecture to use.
At Workato, we understand the risks that CIOs and enterprises face as they bring AI into the core of their business processes and IT strategy. Workato ONE and our Agentic Platform help companies to manage these risks and get real AI use cases into production.
To learn more about Workato’s tools for deploying AI and Agents in your business, contact our team for a discussion about your plans and a live walkthrough of the platform.
