AI-powered automation masterclass: from trends to tangible results

Hannah Colley00:01

Hello everyone, and welcome to today's masterclass on AI-powered automation. I'm Hannah, digital marketing and events specialist here at FlowForma, and I'll be your host for today's event.

Hannah Colley00:14

Before we get started, a few tips for participating. The masterclass is being recorded, and if you need any assistance you can email info@flowforma.com. Your lines are muted, but we want this to be as interactive as possible, so there's a chat box and a questions box on the toolbar. We'll have a live Q&A, polls, and a survey at the end.

Hannah Colley00:52

Let's meet our experts. Gerard Newman is chief technology officer at FlowForma. Gerard leads our product roadmap and innovation strategy, driving advancements in AI and process automation. Hi, Gerard, thanks so much for joining us.

Gerard Newman01:10

Hi, everybody.

Hannah Colley01:12

We also have Niamh Lordan, head of marketing at FlowForma. Niamh is passionate about empowering business people to digitalize and automate at speed, and she champions the message that with the right tools, anyone can drive digital transformation.

Niamh Lordan01:31

Hi, Hannah. Hi, everyone. It's good to have you here.

Niamh Lordan01:42

Before we get into our deep dive on AI automation, I know some people on the webinar may not be as familiar with FlowForma as others, so I'll quickly introduce the platform. FlowForma is a powerful and intuitive platform for digitalizing business processes, from everyday processes to more complicated ones.

Niamh Lordan02:23

From day one, FlowForma has been about empowering both business leaders and IT leaders to digitalize at speed. We were first to market with a true no-code automation solution, and customers love that it's a complete platform: forms, workflow, document generation and insights in one place.

Niamh Lordan02:53

The last two years have been exciting. Gerard's team has been busy in research and development adding an AI layer, which we call our AI suite, on top of the FlowForma automation platform. That's the highlight of today's session, and you'll get a flavor of FlowForma Copilot and our agents and assistants.

Niamh Lordan03:29

We have attendees from across the globe and from different sectors and departments, from transformation and IT to compliance. FlowForma has been used to digitalize an A to Z of processes, from incident management in hospitals to grant applications in public administration and policy management in insurance.

Niamh Lordan04:18

When Gerard, Hannah and I planned this masterclass, it was important to understand your goals for the next six to 12 months, so we included a question on the registration form. Thank you to everyone who filled it in. Gerard, we weren't surprised by the answers.

Gerard Newman04:56

Having worked in process automation for so long, it really is all about efficiency and effective processes. What's interesting is that with the advent of AI, people are looking at how they can use it to make those processes even better.

Niamh Lordan05:08

A few trends came out of what everyone wanted to learn. Number one is increasing operational efficiency, which is our bread and butter. Several people said they're interested in AI but don't know where to apply it in their processes, or what they can automate with AI, so Gerard will cover that with a demo.

Niamh Lordan05:46

And our favorite trend right now is agentic AI. People want to learn how agents can support their staff so they can run things more efficiently and free up more time to spend with customers. Over the next 40 minutes, we'll share demonstrations, tips and advice to help with these goals.

Niamh Lordan06:20

We're at a very exciting point in automation at the end of 2025, and Forrester sums up where we are on this journey very well. Organizations aren't just digitalizing anymore; they're transforming processes with AI at the core. There's been a lot of hype around AI for a few years, but what Gerard and I are seeing are practical use cases and real outcomes in our community every day.

Niamh Lordan07:10

The huge shift is that you can take your process ideas and almost instantly automate them through conversations with AI. I'm not a highly technical person and I don't have a coding background, so that lets me automate faster than ever before. Our community is also automating unstructured tasks that weren't really possible before, and with agents they are now.

Niamh Lordan07:50

We recently ran a webinar with Sarah from the State of Utah, and one of their big outcomes is compliance and audit readiness: when their processes are audited, they get top scores and no scrutiny because they have complete audit records. Customers are also getting better business and process data from what they've automated. So today we want to move past the hype and talk about practical use cases.

Niamh Lordan08:35

Gerard, here's a question I hear at a lot of automation events. Everyone wants to be more efficient, have agents supporting their processes and get better insights, but they don't know which processes in their business are ripe for AI-infused automation. What would you suggest they look for?

Gerard Newman09:12

From the very beginnings of process automation, you looked for processes with high volumes of repetitive tasks, and that's still a good place to start. AI takes it a step further, and agents make it much easier to automate specific tasks. So look at processes with reasonably high volumes of work going through them, made up of repetitive or straightforward tasks.

Gerard Newman09:54

What AI has really added is the ability to support decisions: make recommendations, summarize a process to date, and present information so you can make a decision or continue with an AI-generated one. I'll show a couple of those later, and some of the ways we've used agents to solve particular problems.

Niamh Lordan10:18

Those examples, like credit approvals and risk management, are everyday processes that are very resource-intensive. If AI can help with them, the outcomes can be massive.

Gerard Newman10:34

Absolutely. And it's not necessarily about completely automating those processes. It's about preparing the work and giving it to people to review. That makes the whole process much quicker, allows greater volumes of work to go through, and lets people focus on more value-added tasks.

Niamh Lordan11:03

Next, a poll, because we want to encourage interaction. To understand where everyone is with their AI automation strategy, there are four options: are you identifying process candidates for AI automation, already piloting AI on some processes, scaling successfully, or haven't you started and are still researching your next move?

Niamh Lordan12:10

It looks like we have a clear winner, Gerard, and it matches what we were anticipating. Do you want to share your thoughts?

Gerard Newman12:21

Later on I'll talk about running pilots with AI. I think it's exactly the right thing to do, to try to identify what AI can actually do for your organization, and it's worth looking around to see where it can be effective and start adding value. I'm not surprised some people haven't started yet. Most people are using tools like ChatGPT or Copilot, but the question is how you build them into your business processes so they're used the same way every time and produce more consistent results.

Gerard Newman13:32

About two years ago we started experimenting with generative AI, and we were very enthusiastic to see what it could do. We'd looked at other AI techniques in the past, but they didn't quite suit us. Just like the poll options, we started with a lot of experiments and proofs of concept to explore what AI could do and where in process automation it could really add value.

Gerard Newman14:12

We discovered the strengths and weaknesses of generative AI, and the models have become far more capable since those early versions. This approach of running experiments worked really well for us, and I'd highly recommend it. Even if the experiments are disposable, you're always learning something and advancing your understanding of how to use AI in your own organization.

Gerard Newman14:47

We zoomed in on three areas. First, how AI could help organizations innovate and accelerate their process automation journey, because at times automation is a bit of a slog. Second, how to automate recurring tasks, which is where we started looking at agents. Our thinking from the start was that AI agents would be part of a blended workforce, taking on some tasks and assisting people in their roles. Third, the extra features we could add to support process automation, such as AI summarization of everything that's happened in a form to date, so people can make approval decisions more easily.

Gerard Newman15:54

Here's an example of how we've used AI to innovate. This is our AI-based process builder, and it's available on flowforma.com, so you can try it from our web page. We've invested a lot in it, and people who used it a while ago will see that the features have changed, because we keep updating it as we learn and get feedback.

Gerard Newman16:41

In my scenario, I've been asked to build a data subject access request process. It starts with a form, which I have a copy of, but I don't know much else about the process. The form asks for the requester's details and the details of their request. I load the form into the process builder with a simple instruction to convert it into a single-step process. AI reads the form, identifies all the questions, and creates an electronic version of what was a paper form, with a diagram showing a single-step process.

Gerard Newman17:42

Next, once the request is received, it goes to other departments. I've met the sales team, and they explained that they open Salesforce and check whether there's any information about the subject. So I ask the process builder to add a step based on those notes, and it adds a sales step with questions on whether Salesforce has been checked, whether any data was found, and a summary of it.

Gerard Newman18:46

I'll assume it then goes to marketing, who use HubSpot, and support, who use Zendesk. I type that into the chat, and it rebuilds the process with steps for marketing and support, picking up HubSpot and Zendesk from my instruction. Then it adds a final step for the legal department, and the diagram shows the ability to pass the request back to sales.

Gerard Newman20:10

This is a fairly simple example, and you can try it on our homepage. Without AI, I'd have had to interview sales, marketing and support, document what they need, get their sign-off, and then start building manually. That all takes a lot of time, and the time and number of people involved are a barrier to process automation.

Gerard Newman20:46

You're not just creating a mockup. The same capability is available in Flow Designer, the tool we use to create processes. I load the same documentation with the same instructions, and it creates a working process: I can enter the details of a data subject request, which could be from someone outside the organization, and it's pushed through the steps of the process.

Gerard Newman21:24

This video takes about five and a half minutes, so in five and a half minutes I built a working process for something in my organization that I don't know much about. That's why we were so impressed with generative AI for process building: it has a large knowledge of processes and how people go about their jobs. If you can engage parts of the business with a process that's already built, it's far easier to get engagement and feedback than starting with a blank page.

Niamh Lordan22:20

I love how fast and interactive this is. You can test it yourself on our homepage: upload an image of your diagram or type in the process you want to digitalize. At automation events, people tell me they don't know how to map a process or get started, so this really accelerates the journey from something we need to digitalize to a visual you can show stakeholders and get signed off.

Gerard Newman23:00

Absolutely. We use Copilot extensively ourselves to build processes for demonstrations. One really nice feature is that you can gather around a whiteboard, sketch out a process, take a photo of it and load it into Copilot to start building. That takes something from concept to reality really quickly, and it means you can build processes that are effectively disposable. Running a marketing campaign for four months? Build the process, then get rid of it at the end.

Niamh Lordan23:42

That brings us to our next topic. Registrants wanted to understand how agents could support their staff and make operations slicker, faster and more compliant. It's the hot topic of the moment, so Gerard, can you show us how to instruct these agents and share a few use cases?

Gerard Newman24:17

At the start of the year we made our predictions for 2025, and agentic AI was all over them, so those came true. I'll start with a simple example of creating an agent and getting it to do some work. A single-purpose agent is most successful when it has a clearly defined task and you expect a clear outcome. If you're doing that task many times, it's a good candidate for an agent.

Gerard Newman25:05

Whether you call it programming, training or instructing, you do it in natural language. Give the agent clear instructions on what you expect it to do and how to handle any exceptions, and keep your prompts generic so they aren't tied to one specific type of information. In this example, the prompt tells the agent it's a data entry agent that processes invoices, lists the data to extract, says to return "not supplied" if a value is missing, to ignore currency symbols in amounts, and to record how long the job takes.

Gerard Newman26:01

This is a simple invoice processing flow with three steps: invoices are received, the data is entered on a form, and it's passed to our accounting system. It's deliberately simple, to show a basic agent operation. First, the manual way: here's an invoice from Loom for $18, received in May, and I enter the details by hand.

Gerard Newman26:56

Now I set up an agent to do this for me. I add the prompt from the previous slide, then tell the agent where to put the information it finds: which questions on the form to populate. I load the invoice again and submit it. The agent reads the invoice, extracts the details, and populates every field on the second step. It didn't find a VAT amount, so it returned "not supplied", and it extracted the currency correctly.

Gerard Newman27:52

There's been technology for extracting information from documents for many years, but most of it takes time to implement and configure. This is a really simple way to do it, and AI is very good at extracting information from lots of different document formats.

Gerard Newman28:19

Now some more complex use cases. We built a proof of concept for claims processing in general insurance. Complex tasks are achieved by agents working together. Some agents read policy documents and claim forms and bring the information into a standard format. Others use tools we built to pull data from your existing policy and claims systems. Together they assemble the information.

Gerard Newman29:18

That information goes to a decision-making agent that validates whether the claim should be processed. For example, a claim for a motor accident on a home insurance policy wouldn't be valid. Then the assessor review and risk analysis agents do deeper thinking: they assess the damage claimed, whether it's covered under the policy and its terms and conditions, look at your claims history, and calculate a risk score to flag anything suspicious.

Gerard Newman30:30

All of that feeds a recommendation agent, which gives the claims assessor a recommendation they can agree or disagree with. It saves a huge amount of time, and customers get a faster response. In insurance, the cost and time of claims handling are huge factors, both for managing costs and for customer satisfaction, so anything that makes it faster, more consistent and more automated is valuable.

Niamh Lordan31:27

It makes the process much more efficient, gives employees a better experience with agents supporting their work, and they process claims faster, probably with fewer errors.

Gerard Newman31:45

Absolutely. Data errors are eliminated, and as you develop it you can refine the agents' instructions, which tightens them up and reduces exceptions. But it very much keeps the human in the loop. The goal is to give the assessor a well-prepared piece of work they can pick up, instead of scrambling to gather information from different systems.

Niamh Lordan32:38

The next use case has been requested a few times in the last couple of weeks: transforming tender management. Lots of people have spent long nights working on tenders. How can agents streamline this process?

Gerard Newman33:04

We developed this one after talking to a number of customers and prospects. Tenders are really complex, with huge amounts of documentation in different formats: text, data files and drawings. The objective is not just a response, but a response that wins. We have agents that consume all the tender documents and supporting information, plus information about your organization and your USPs so they can be worked into the responses.

Gerard Newman34:14

With that knowledge base in place, a group of agents is each prompted to answer specific questions in the tender: your approach to delivering the project, the materials, the costings and so on. They produce initial responses that people review and either tailor or use as they are. The tender team can also ask the knowledge base direct questions, which helps with very large tenders, because people aren't great at reading and remembering all that information, but AI generally is.

Gerard Newman35:24

Here the answers come specifically from the knowledge base built on the documentation you provide, combined with an LLM, and your data is kept secure and separate. The thing to watch is hallucinations, but those can be overcome too. Overall, you significantly speed up delivering a quality response, and the tender team can focus on making sure your USPs come through.

Niamh Lordan36:05

That's a nice segue into our third use case. Tender submissions often include security questionnaires and vendor management, which is a huge and important focus for organizations and takes a lot of time. Gerard, can you talk us through our new product, FlowAssure, which launched very recently?

Gerard Newman36:55

This is a very topical issue. Onboarding vendors, particularly IT vendors, is really important for maintaining data security and control across an organization with functions distributed to different vendors. It matters in small organizations, but with thousands of vendors it's a really big issue, and onboarding is becoming more onerous.

Gerard Newman37:26

We've productized this with a team of agents. For application onboarding, there are agents that read vendors' pen tests, interpret completed questionnaires, read ISO reports and read SOC 2 reports. They identify shortcomings or areas that need clarification, generate the questions, and send them to the vendor directly or to a cybersecurity analyst. Non-critical or low-risk vendors can be approved automatically, so analysts can focus on critical applications and significant questions.

Niamh Lordan38:46

Those are three practical, real-world use cases that our customers and prospects are using, well beyond the hype. I'm always thinking about outcomes, and using AI for outcomes, not for its own sake. At FlowForma we're trying to create a world where work flows as it should, approvals happen as they should, and insights are available when and where you need them.

Niamh Lordan39:37

To help you on your AI-infused automation journey, Hannah is launching a quick poll. If you'd like a consultation with one of our AI experts, or you're interested in a pilot or trial, submit your answer and our team will follow up with you directly.

Niamh Lordan40:04

Time is flying, so before Q&A, a few tips for your journey. It's not like flicking a switch and having everything AI-powered. My first tip: don't just add AI. Think about the problem you're trying to solve. Like any project, if you focus on the outcomes, you'll succeed.

Niamh Lordan40:46

Analysts such as Forrester and Gartner stress that to power AI in your processes, you need a solid, repeatable process foundation. AI can amplify good processes, but it won't fix bad ones. That's where Copilot helps, by automating those processes so you can add agents on top, with data validation built in, which gives your AI better data to work from.

Gerard Newman41:30

Absolutely. In many cases, that's what AI is about: using clean information from a good source.

Niamh Lordan41:41

We're all about keeping humans in the loop. AI and agents are fantastic for predictions, suggestions, commentary and moving processes in a certain direction, but we need people to make the smart decisions. And rolling out AI isn't just a technology change; it's a culture change. The customers who do it well communicate with their staff, train them, get feedback and bring them on the journey.

Niamh Lordan42:31

Like any good project: test, learn, deploy and test again, which is so easy now with FlowForma Copilot, because you can prototype processes quickly and get feedback. Don't let perfection be the enemy of progress. Get your process digitalized, add AI this week or next, and keep improving. Someone in the chat asked about continuous improvement, and that's exactly the tactic.

Gerard Newman43:15

I totally agree. Traditionally, process automation often started with large analysis projects to understand every process before doing anything, but those days are gone because processes change so quickly. It's far better to start with something, automate it, put it in use, get feedback and change it.

Hannah Colley43:50

Now for our live Q&A. We have a good number of questions, and if we don't get to yours, we'll follow up with you. First: we have lots of process maps that need automating. What file types does FlowForma Copilot support?

Niamh Lordan44:23

If you have your processes mapped, you can upload the diagrams as PNG or JPEG files into Copilot, and it will transform them within seconds.

Hannah Colley44:46

Someone says they love the agent examples. How much training do these agents need?

Gerard Newman44:56

It's not so much training they need, because they're based on generative AI and the LLM behind it. What they need is clear instructions: tell the agent what you want it to do, what you expect the result to be, and how to handle unexpected events, usually in plain English. As with any programming, you might not get it right first time, so tweak the instruction and test it with different examples to make sure it gives you the results you want.

Hannah Colley45:36

Are there any prerequisites for using the FlowForma platform?

Niamh Lordan45:46

FlowForma is an app that sits on top of SharePoint Online. If you already have Office 365, in nearly 99% of cases you'll have SharePoint Online, so that's the only prerequisite.

Hannah Colley46:01

How do you track what AI has done or suggested for audit purposes?

Gerard Newman46:09

That's an important question, especially for agents working in a process. You need to understand what they've done and the quality of their output. This is music to our ears, because we believe a good place for agents is within your existing processes or a structured process environment. There, they work like a person would and are subject to the same measurements: you can see the work they took in and the output they produced, and it all flows into your Insights dashboards so you can monitor how they're performing and how much work they're doing.

Hannah Colley46:56

Are there limits to the process complexity, size or frequency that FlowForma AI can handle?

Gerard Newman47:07

There are no real limits. You can build complex processes with Copilot, and feed in your diagrams as you saw at the start. It's focused on business processes, so it won't build processes for baking a cake, and it isn't designed for very technical integration processes that tie systems together through a set of steps. Other products do that. It's designed for the business processes organizations use to get work done.

Hannah Colley47:52

We're planning our migration away from InfoPath, and this seems like a good alternative. Is this a common use case for FlowForma?

Niamh Lordan48:05

Yes, definitely. We recently interviewed a customer, Morley College, who have just been on this journey with us. Copilot makes migration easier than ever: we can take captures of your InfoPath forms and upload them into our product, so migrating to FlowForma is straightforward.

Hannah Colley48:37

We'll finish the questions there and get back to anyone else. For next steps, you can book a complimentary demo tailored to your processes, and don't forget you can head to the FlowForma playground on our homepage right now to start prototyping your processes. Thank you so much for joining us, and thanks to Niamh and Gerard.