Clip
Summary
Andrei walked through six real-world patterns for Evo Workflow Automation, then showed one in action. A scheduled flow pulls Salesforce cases from Databricks, hands them to three agents — themes, risk flags, synthesis — and emails the team a weekly digest that doesn't just summarise but recommends what to act on. Old workflow tools gave you a report. This gives you a plan. Inspired by Chris Joel from the field.
Screenshots
Transcript
That was it for my demo. I'll hand it over to Andre.
Andrei Pascu 48:29
Thank you very much, Alex. So, yeah, I would like to show a short recording of Eva workflow, how it's used in or how it's planned to be used in real life usage, and I would like to thanks to thank Johnny Ivat for this, and... The Chris Joel, the Director of Professional Services Transformation, so...
I would like to share my screen, and... Start the recording. I come up with six ways that we or customers might use Evo Workflow.
So the first one is systems integration. So we might be moving data between two systems. Evo workflow makes it a lot easier for a non-developer to integrate 2 systems.
The systems might have API endpoints, or maybe their systems have got flat files, and we've got to move those files between systems, download them from one system. input them into another system, et cetera. So I think if a workflow gives us that capability or gives non-developers that capability. I think workflow can be used as a way of extending our own products.
So within the product you work on, you know, you might be looking for an opportunity to add a workflow capability orchestrate a process and you could obviously send that process out into the workflow. Workflow can handle the orchestrate the routing of. of it and then it can come back into the product that you work within. The third, I suppose, pattern or example is around proactive monitoring or scheduled workflows.
That could be a sort of monitoring workflow, so on a scheduled check condition and take some actions. Or it could be waking up every day and going getting a list of data out of another system. you know, sending that in a digest or form to a user. So lots of examples where we might want to do things on a scheduled or time basis.
We can obviously involve AI in those processes, or it could be entirely deterministic. Fourth one could be off-system approvals, so... Customers might have an approval process or request the process that doesn't exist within one of the access products.
Lots of things that we have in the access group are sent and sent round and managed by things like Microsoft Forms, right, or even Excel spreadsheets being sent via email. Why not create the capability to have those processes. handled and orchestrated by workflow system, whether that's a new supplier request or a bonus approval, whatever it is, we can create a simple form via micro front end and use EVO workflow to route that approval around the organization. Then the 5th and 6th ones might be a little bit more interesting.
I think we can use either workflow to create tools for agents. It could be an entirely deterministic flow, quite a complicated flow involving multiple steps. We can, if you like, package that up as a single operation, a single callable tool. in the same way you might do with MCP server, but the agent can call it and get one back, one clean result back.
And in that final example, multi-agent orchestration, either workflow allows us to put together multiple agents, each agent doing a small specialist task, a little bit more reliable, and then use Evo Workflow to make sure we move the work or the data within those different agents. So those are my ideas of kind of six patterns or six ways you can use Evo Workflow. And now we're going to have a look at a couple of things I've built that have got some of these capabilities.
And so, yeah, you can see what we've been doing with either workflow. So, my first example is a workflow which runs on a schedule basis, so running actually every day. So that's what the schedule trigger is doing on the left-hand side.
It goes and gets some data from Databricks, it builds some email cards, and it sends out to a group of users any recently closed one sales opportunities. Okay, so kind of like a daily digest. It looks back to the last seven days. and sends out an email.
Nice simple example. So, fairly simple. And I guess, you know, we've had workflow type software for quite a while.
You know, we've been a reseller of some workflow software. People involved with the ERP soft codes be aware of what used to be called Orbis Task Centre, I think it's just called Task Centre these days, a product that we've resold for many years. that does this right, goes to a database, queries the database, creates a report. And so then I changed myself.
I said, well, okay, but what's changed now? We've got workflow now, we've got AI, how could this be smarter? And so I came up with a different idea.
Similar principle where I would go to Databricks, I would pull out some Salesforce data, in this case Salesforce cases. But the thing that's changed now is that rather than simply send the user a report, which is open to interpretation, the human has to read and interpret and then plan what to do next, what I've done instead is I've got three agents involved. I've got a theme analysis agent, A risk agent and a synthesis agent that basically rolls everything up into one summary as well.
O, when this workflow runs. I process those cases, I group them by type, so the different case types. I then condense some of the data down for analysis and we pass the first agent, the theme agent, the theme analysis agent.
It gets a truck to Jason. which are grouped by case type. And its job is to identify the reoccurring themes within each case type group. And then, obviously, explanation and series of rules.
The next agent, the risks flags agent. gets an input from the previous agent. But also gets the original case data. And this agent has a different job, to identify operational risks.
Not to repeat the themes, but to flag cases or patterns that suggest something needs attention. And then an example of signals to look out for. And the final agent, the synthesis agent, takes all that data.
So I get the risk flags data. And its job is to focus on what needs action. And create a set of recommendations.
So 3 agents working together. Multi-agent workflows, talked about that in my themes at the beginning. We built some email cards, email template, Microsoft Outlook, and what we get as a result running weekly is one example.
A weekly case digest. that isn't just a summary of what's been happening via Salesforce cases. But it also has some analysis provided by AI. Of critical patterns.
Suggested actions, risk flags. Things and statuses, flags with high or medium. And there they are, see his recommendations down there as well.
So that was it. Hopefully this was inspiring for you and we are looking forward to see you using Evo workflow automation.