Improving AI Workflow Tools in Fibery

[Yuri_BC]

Fibery has huge potential to become a powerhouse for advanced data analysis by pulling in data from different sources and enabling AI-powered workflows across multiple layers and processes.

This year, I’ve been experimenting with chaining AI operations in Fibery to connect and process data in smarter ways. The challenge, though, is that Fibery’s current automation features haven’t kept pace with AI capabilities, and it’s starting to feel like a bottleneck for the product’s growth.

Fibery really needs a dedicated interface for building layered and connected AI workflows. The Automations UI just doesn’t cut it—it’s not built to handle the complexity of chaining AI processes or making it easy to mix and match operations to get the most out of all the data in the system.

Right now, it feels like only the most tech-savvy Fibery users can figure out how to set up these kinds of workflows. But this is where the real opportunity lies: making advanced AI workflows simple and accessible for everyone, not just for admins or developers. That should be Fibery’s main focus moving forward.

There’s also a gap between Fibery Templates and the Automations system:

• Templates are great for creating basic, standalone setups but don’t work well if you’re trying to integrate them into existing structures. They feel more like static examples than reusable building blocks or plugins.
• Automations are powerful but are geared toward advanced users with the time and skills to set up complex workflows.

The real trend in AI right now is giving everyday users the tools to create layered AI workflows and insights with minimal effort. Fibery needs to lean into this trend by building tools that make AI integration seamless and intuitive.

Personally, I’ve never used Fibery Templates and probably never will. They just don’t plug into my workflows in a way that feels dynamic or useful. What Fibery really needs is to close the gap between static templates and overly complex automations, making AI-powered processes a core, user-friendly feature for everyone.

[calh-fsp]

I think @Yuri_BC 's points about Fibery’s AI workflow gaps are spot on. The current one-off automations serve their purpose for individual tasks, but there’s a significant opportunity to unlock more advanced functionality by harnessing Fibery’s flexible data model. Instead of just relying on AI chatbots to surface data, imagine introducing configurable Agents capable of executing complex, multi-step workflows—creating new entities, updating existing ones, and intelligently interacting with data across different spaces and databases. Platforms like Copy.ai have already illustrated the power of chaining AI-driven tools together (e.g., “generate blog outline” → “expand sections” → “draft social posts”) to create seamless workflows. Fibery could take this concept further by leveraging its well-structured data model to enable something similar.

Imagine having the ability to create a Marketing Agent, a Support Agent, and a Product Manager Agent, QA Agent, etc., that can be pieced together into an agent-driven workflow: Input a description of a feature → Create PRD → Generate Epic/Story tickets and assign to Dev → Create Test Cases → Generate Help Center Article, and so on. These agents could not only automate repetitive tasks but also collaborate across different parts of the organization, ensuring consistency and efficiency while reducing manual hand-offs.

This feels increasingly relevant as agent frameworks like LangChain, AutoGen, and CrewAI continue to gain traction. These tools demonstrate that when AI agents are provided with strong contextual understanding and access to structured data—both of which are inherent strengths of Fibery but with no-code—they can manage sophisticated, multi-step processes with efficiency. Drawing inspiration from solutions like Copy.ai and incorporating an visual workflow builder, Fibery could evolve beyond simple automations and static templates to deliver intelligent, end-to-end process automation. This would be a game-changer for teams aiming to systematize workflows while retaining the flexibility that makes Fibery stand out.

[calh-fsp]

I can’t help but continue to dream how powerful AI Agents & Workflows could level up Fibery 1,000x. In hope of some community feedback or further collaboration, I’m sharing some initial progress I’ve made while tinkering with implementing an Agentic Workflow workspace in Fibery, albeit with all of its limitations. I’ve published this public Fibery whiteboard with my current thought process and concept progress.

KEY CONCEPTS

Workflow as an entity

Agentic Workflow entity represents a collection of inputs, steps, & outputs
Whiteboard linked to entity to visually build the workflow.

Input Blocks in Whiteboard

Elements on the whiteboard that may be used as “inputs” into downstream elements.
Would be nice to consider “Input Block” field types on the entity (or at least expose current entity fields as input options that can be displayed independently on the whiteboard).

AI Blocks in Whiteboard

Elements on the whiteboard that can receive inputs, a prompt, run an LLM command, and return a result.
Blocks could be an AI Action (direct call to an LLM) or AI Agent entity (see below).

AI Agent Entities as AI Blocks

As illustrated in the diagram above, I have an AI Agents database with fields: System prompt and User prompt (the user prompt accepts an input from a content block or previous workflow step).
Would be nice to consider “AI Agent” field type (or relation) on the entity. Given the user has added “AI Agents” to the entity in the entity page/form, when they navigate to the Whiteboard, those AI block elements will automatically be added to the whiteboard.

Historical Tracking of Workflow Runs & Agent Responses
I’ve created Agent Responses database, which tracks historical “runs” of an AI Agent entity. It’s basically inherits the AI Agent class, copying the System & User prompt at that point in time and the resulting “Response” for that agent run. This will help with viewing AI Agent efficacy over time and any improvements/degradations based on varying system/user prompts.

I’ve also created a concept “Workflow Runs” database to do the same, track historical completions of an Agentic Workflow run. This will help with viewing changes to the workflow over time–inputs, AI Agents/Actions, etc.

---

That’s it for now. I’ll be continuing this concept to see how far I can get. Hope this sparks some good conversation or interest in collborating to come up with a nifty Agent solution in Fibery.

[Yuri_BC]

Thanks for sharing that, Cal. You did some very useful prototyping.
I did a similar thing, a bit different but also workflows with AI agents. Maybe we can join forces here to test out what we can accomplish in fibery as it is now. I’ll post it one of these days here.

[mdubakov]

@calh-fsp

I can’t help but continue to dream how powerful AI Agents & Workflows could level up Fibery 1,000x. I

We are working on Agents right now, but our vision is just code. Our agent writes and executes code to complete the workflow. As a first step, we will release query agent soon (it can handle pretty hard questions about your data in Fibery), and then we are working on A2A agent implementation already that will connect various Fibery agents together.

You can play with Q&A agent at your workspace, just use this URL and replace xxx with your Fibery workspace name https://XXX.fibery.io/api/ai-answer/thinking-agent/thinking-agent.html You can try ANY question around Fibery data you have.

It would be interesting to hear your feedback. Note that the agent is quite slow and almost always ask for clarification on the first step. Anyway, results are usually quite good.

As an example, here are few prompts we use to check how good it works

Find feedback from September 2024 till March 2025 from our top 10 largest workspaces and identify problems that were reported by these workspaces. 

or more detailed version

Find feedback from September 2024 till March 2025 from our top 10 largest workspaces and identify problems that were reported by these workspaces. Approach this in a structured way:
1. First, query the Crm/Workspace database to find the top 10 workspaces with the highest "Weekly Active Users" value.
2. For each of these workspaces, fetch all highlights from the highlights/Highlight database where:
"Source Workspace" matches the workspace
“Source Date" falls between September 1, 2024 and March 21, 2025.
3. extract the content from these highlights using the Highlight Snippet field
analyze this feedback to identify problems (find  at least 10 problems for every workspace) and exact quotes for these problems
4. Finally, present findings for every workspace like this:
Workspace Name
Total Highlights for this workspace
List of Problems with exact quotes for this workspace 

or another example

Which product areas get the most feedback in January-February 2025? 
Plan:
1. Gather feedback data from highlights/Highlight. Do not query other databases.
2. Filter the data to the January-February 2025 timeframe. Use "Source Date" for filtering the timeframe
3. Use highlights/Target Product Area (as a link to Product Area)
4. Link each feedback item to its corresponding Product Area
5. Count and aggregate the feedback by Product Area
Provide results as a markdown table with columns: Product Area, Highlights count

[Yuri_BC]

This is very promising, thank you @mdubakov. Can we give feedback on that Fibery Thinking Agent in this thread, or a seperate thread?