Skip to content
Back to videos

Video

80 Hours Down to 2: AI Is Rewriting Business Workflows

Matthew Sutton, founder of the Salesforce implementation partner Resonant 360, explains how AI agents fit into business workflows and how one development task fell from 80 hours to about 2. He and Pete Cooper also cover moving a proof of concept from n8n into Agentforce, data sovereignty, AI redundancy and what work could look like in 18 months.

Video summary

Matthew Sutton founded Resonant 360, a Salesforce implementation partner, in 2015 or 2016 and says the firm has worked on more than 400 implementations. He describes Salesforce as a business platform for the customer-facing side of a company, covering sales, service and operations, and says what sets it apart is how much automation it allows. He says he used to describe AI as "automation with fuzzy logic", where a process can interpret a piece of data and make a decision instead of needing an exact rule.

He came to Salesforce as a sales manager around 2013. He says an out-of-the-box rollout at that company made the processes so clunky that his sales revenue dropped by about 50% from go-live, so he got an admin login, changed the code in a sandbox over a weekend and was hooked. Back then the goal was fewer clicks and pages. Today, he says, the question is what work should look like. A user interface requirement that would have taken one of his developers 80 hours came down to 8 in January 2026, and he says he now does it himself in about two hours on another screen at his desk.

He traces a longer arc too. In 2020 a customer moved from another CRM to Salesforce and, by his account, the same team handled five times the cases. In a more recent project his team built a proof of concept in n8n that found a purchase order number in an email thread and attached it to the CRM. The customer's security team was not comfortable with it, so they moved it into Agentforce, Salesforce's agent capability, once its cost had come down. He says Salesforce's trust layer anonymises data before it reaches an LLM, and that his firm can now use AI hosted in Sydney so data does not leave the country, which matters for clients such as financial services firms. He also argues for redundancy, after a bug in one provider's credit purchase process left his team out of tokens and unable to work.

Matthew thinks most businesses are asking how to apply today's AI when they should be reimagining what their work will look like in 12 to 18 months. His own team records customer conversations and uses Claude and Codex to turn them into requirements documents and human-reviewed scopes, and he estimates that 80% of their tasks could be automated. That creates a new interface problem, because people have to keep up with far more work, so he is building software where the agent changes the screen to match the conversation. On downsides, he expects a shift in the workforce and points to IKEA, which he recalls retrained customer service staff as stylist consultants rather than letting them go.

Asked whether AI is hype or help, he splits it. The hype is that nobody knows what the future holds. The help is what the tools can do now, taking away the parts of the job people do not like, such as data entry, so a CRM can be a sales and customer experience tool rather than a reporting tool.

Video transcript

Read the spoken content without loading the YouTube player.

We are sharing a lot of information with the foundations. The the use of Claude is is is has exploded. I mean, it's flattening out now, but >> What is your relationship with Salesforce? Do you act as an agent for Salesforce? Are you employed by Salesforce? >> The capabilities of AI in the past 12 months has exploded. >> So when when when would you going back to to saying it was expensive? When did you experience that? >> I was excited in January this year when I could put a requirement for a a user interface to my one of my development team. And something that would have taken 80 hours came down to 8 hours in January this year.

Where we are now though is 2 hours and it's something that I'm doing on another screen on my desk. >> AI, is it hype or help? Good day and welcome to the show. I'm your host Pete Cooper and our topic is AI hype or help. Today's guest is Matthew Sutton from Resonant 360. Matthew comes to us from the Central Coast of New South Wales, Australia, somewhere I know well. It was where I grew up. But he's going to talk to us today about Salesforce and how that is revolutionizing the way businesses work. Salesforce is used to be a CRM. It's a little bit more than that now.

And he's talking about massive changes in efficiency and capability with this product thanks to AI and his implementations, particularly in workflows in businesses. This is very interesting, very grounded, very practical. Let's get into it. Good day Matthew. Welcome to the show. How are you doing today? >> Uh very well, thank you. It's great to be here. >> Yeah, thanks. Where are you coming in from? >> Uh currently Central Coast, New South Wales. >> Yep, that's Australia for all those people that don't know. Yes. Yeah, so I grew up around that that part of the woods too. How long have you been there? >> Uh about >> months I moved out to Sydney. >> Yeah. >> So, yeah, relatively new. >> Yeah.

Well, most people know where Sydney is. Most people would say, why would you move out of Sydney? Isn't it a nice place? >> Um, space, busy. So, [snorts] change of pace, I think, really. >> Mm, mm. And you're happy up there now? >> Very. >> Yeah. Well, Matthew, what do you do? Tell us a little bit about your business. >> So, I founded uh Resonant in 2015 2016. So, um we are we've been a Salesforce implementation partner looking at automation and business processes and implementation since then. Um, and it's about it's been business efficiency and and it process improvement ever since. Um, so it's working we've we've worked on over 400 implementations in that time.

We we've got clients we've worked with for over 10 years. So, focusing on been uh continuous business improvements. >> Now, Salesforce, is that a CRM? Is it a platform? Just just break it down in layman's terms, what is Salesforce? >> The Salesforce is uh a business platform that focuses on streamlining your processes and uh capturing the data that's needed at each stage. So, it's known as the um the founder of SaaS started with Sales Cloud. So, it did focus and and initially on sales, but moved across the years we're now sales, service, and operations. So, it's the whole of the front end, the customer facing pieces of the business.

How do we uh engage new customers and how do we manage and look after a new customer um and then deliver services to them? It's different to an ERP, which is about the back end of the operational side. The Salesforce operations is about the uh combining that with the customer experience. So, it's the it is the the customer 360. >> Okay, good. So it's it's about the interface between your clients company and their clients and making sure that works I guess efficiently gives the client what they need and I mean is there any company that can use Salesforce or is it just a particular type?

>> Um it can be used for any company we've worked with we worked worked with companies who are one or two people and looking to capture and automate at the efficiency at a very early stage and we've seen those scale into multi you know multiple times over those years but also with enterprise clients as well. So we work with customers global organizations who are well known and common names. >> So if I had a CRM and I've worked with a few over the years like Zoho, Brevo, things like that which to me kind of just help you capture it's like a front end of a database I always think of the CRM not particularly clever and but it'll do reports and things like that.

So so what beyond the CRM does this new Salesforce have? >> So I think Salesforce has been different throughout the past 20 years by the level of automation that's capable. So back when we started or when I started 2013 2014 we had process builders. So when when this happens do that. So it was about automation at that time and I was lucky enough to on a project to get involved before I started this business get involved on the global rollout of flows which elevated the level of automation into logic trees but without code so it was no code automation. >> Mm.

>> That's progressed over the years into um a a framework of automation which isn't available in other platforms and other platforms are only starting to catch up with that I would say in the past two years and a lot of them don't. If this do that has been coming slowly over the to all platforms so everything has that but the the ability to create completely custom automization experiences using flows is something that's been distinct within Salesforce and now as we move into or we have moved into the agentic era it's the I used to explain the AI would be it's automation with fuzzy logic. >> Mm-hmm.

>> So automation used to be a distinct it has to be exact and we're able now to have some interpretation and fuzzy logic within that. >> It's funny you say fuzzy logic that I haven't heard that term since I think I must have been 20 or 30 years ago and I can't remember the context exactly but from memory I think it was around getting the focus right in cameras because I was in cameras for a while so that would use fuzzy logic to to figure out what's the correct focus is because the correct focus in one part of the image isn't necessarily the correct focus in another part of the image and you have to kind of [clears throat] use fudgic logic to kind of do the best you can. >> Yeah.

>> Kind of like a precursor to AI I suppose. >> Well that that's it and that that's how I used to when it when three or four years ago as we started using agents in Salesforce to interpret a piece of data and make a decision about it during a distinct process. >> Mm. >> The way of the way of kind of communicating that was yeah fuzzy logic. >> No I like it. I mean I think fuzzy logic automation with fuzzy logic I'm I'm to use that again. So what is we're going get into the AI stuff in a sec, but just so I got a proper grounding here, what is your relationship with Salesforce? Do you act as an agent for Salesforce? Are you employed by Salesforce? I mean, how do they?

>> Uh, so we're we're an implementation partner. So we work with customers to get the the best use of the tool and upskill them on the possibilities and the art of the possible and try and lead customers with the ability to take ownership of that journey um and continuously improve their their business processes. >> So do you find work for Salesforce? Does does does Salesforce bring you on as an implementation partner when they've got a client? How do they normally? >> I think it's been a mix of both over the years. So there's there's referrals for um from customer to customer and there's that Salesforce uh luckily we have a you know a great track record.

So there's a quite a close close partnership in that side. >> Yeah. And then what's the can you crystallize for me what's the difference in terms of a business return on investment or what would a business owner say yep I need Salesforce Salesforce versus a regular CRM? Why would someone want to go that far with Salesforce? >> I think we see customers migrating to Salesforce from other platforms because those platforms are not able to be flexible enough to meet their business processes. >> Mm.

>> Uh you know we can build so many different processes and systems within Salesforce that we can meet virtually every business requirement um which is a blessing but also a curse because it means when we have we are asked a question can we do this the answer is invariably yes we can but should we and how how much customization is that going to take? So the question would be should we then do that? Is it the right choice? Will we be able to see a return on investment?

>> Yeah, so if you're asked to do some feature that you don't have already built, then is it generally incumbent on the client to pay for that or is it something that you internally decide with Salesforce just we're going to develop that feature because we can see it going into other >> Yeah. And no, it's it's 100% it's the customer's decision because it's a custom solution for them for their IP to retain their personal business processes and Salesforce have no um distinct um access to that IP. So it is a way of protecting your own the way that your business runs and works on that IP. >> Right. So you said sales, service, and operations. Did I get those three right? >> Yes.

>> So I'm going to think of a little example where a company makes widgets. Um they somebody calls in, they say I want to buy one of your widgets. That's the sales part, right? That's collected somehow, goes into the Salesforce, is handled by a person or maybe not. Do you want to talk about that in a minute? Then somebody has to ship that, right?

And so that goes over to operations and they pick up the order, they process it, they box it up, they ship it out and it's again updated in the system and then they're they're using it and a while later it needs regular service or there's a problem, then they they call in or they send it in and it's pulled back up and there's a service element done and then it ships back out. Sales, operations, service. Did I cover it well enough? >> That's it. So we we would use the the sales process for quoting um and pricing and delivering the information the customer needs, making sure we're not missing out on opportunities or the pipeline so we can track the pipeline through.

Things aren't lost in boxes. When the sale is made, we can then generate um assets and unique assets with uh reference numbers. That would square the the connection with an ERP would would make sense because we're not changing the warehouse uh back end, but we the one view of the customer is to have the visibility because then we can update the customer on an order from from the CRM. Um and then it moves into um service. So yeah, we would create a case in other systems it's called tickets um linked to the asset. We know exactly when that is. So it's the one view of all the information. Um so when a customer rings up, the person they're speaking to knows the history.

>> So why would someone pick Salesforce? I mean they've been around for ages. I've known them for a long time. Why would they pick Salesforce over a competitive brand? What differentiates Salesforce? >> I think it's the maturity of um of the business process that that that's trying to be achieved. It's about having your business process IP embedded into your CRM, so you are able to differentiate yourself. Um if we look at um there are smaller industries where bespoke packages means that everyone who's using the same package is operating in the same vein. There's no differentiation in their service because the the systems that they're using are black and white.

What we what you want to be able to do is differentiate yourself from your competitors by having your brand within the service and the customer experience. >> Yeah, okay. Okay. That sounds great. So um would so yeah, Salesforce sounds great. How long you've been doing this now um Matthew? >> Um so since about 2013 is when I first got my hands on Salesforce. >> Mhm. >> Um and it's been an evolution ever since that time. >> Yep. Yep. Did that was that that bring you to Australia? >> No, I was here uh at the time. I was working as a sales manager um and the the company uh implemented Salesforce that time.

Um the implementation was out of the box essentially, and but the business processes became quite clunky, uh difficult to find. Essentially my sales revenue dropped by about 50% from go live. So and we had a few months, we had about two or three months of trying to work around those problems with the partners we were working at the time. Um but after about three or four months it was we had I had to do something about it. So I persuaded my way to to get in involved in the project team. I got an admin login to the system on a Friday and there was a piece of work there was a piece particularly that was clunky for the users to actually try and run their day-to-day.

Um the the feedback we had was that's that's going to be a big change. It's going to take a long time. It's two or three weeks of coding. And I and I I knew at that time that I hadn't read code for about 15 years, but I I could see a better way of doing it. So I got the admin login on a Friday in the sandbox, logged in on a Friday night, found the code and read the code and then made the code made the change to the code over the weekend and came in on a Monday and said I've done it. So it was at that point that I got hooked I think at that point and I could see the better way of doing things. And the progression was you know I wasn't a coder. >> Mm.

>> It was in a sandbox so it was safe but it was proving a point. Um and that but that's when the the Salesforce account exec at the time Tom Green [clears throat] so thanks to him um got us onto the global pilot for flows. So flows at that time were the new way of doing automation where we could have logic trees and decisions. Um and I think that change that that process of changing it was distinct of when when I came in and and focused on business process. I think over the course of that year. >> ask you. I mean in this this AI world today, was this before Claude? Was this before, you know, the >> This was yeah, this was absolutely before Claude.

Um and the the difference now looking back on it is is it's incredible. The so I flows at that time they were brand new. They didn't have any error handling. If it didn't work, you had to kind of really just sit there and figure it out because the visibility of what was going on with the automation was was tiny. So I every every time there was a piece in the flow, I had to add automation for it it to email me every piece of the logic. So I could actually track it. Now there are multiple um notifications about errors and pieces in the screen and if it if there is an error, the flow error that you get emailed is so detailed. It tells you exactly what's going wrong.

But back then it was incredibly slow, but I mean I within the course of the first year I we built I I [snorts] I built over 40 pieces of of logic. So I was turning out business logic pretty quickly at that time using this brand this new tool um to to change business processes and get efficiency back into it so people had what they needed in front of them. So that that the difference between then and now when we look at that automation progression of the past 13 years is but then it was about too many clicks, too many how do I you know make the the user empower the user to do their job as efficiently as it can. But the question today that I'm asking is what does what should work look like?

So it's actually evolved from just trying to make thing people move quicker and we've done that progressively over the years. Now it's thinking about, well how should they be working? >> Yeah, 100% >> that they're doing. >> That idea of the clicks, I remember, you know, many software tools that I've had over the years that we we used to say it stares back at you, you know, it's just a screen and then you've got to go click.

I mean just up when I was doing some sales work, you had to open the app, type in the first name of the person, then there'll be a whole list of people with the first name first name, you scroll down, click on that, then you'd have to go over to another field where you wanted to just make an update, click on that, select a menu, then it would open up and you'd write left a voicemail. All that work just to write that. It was so painful. That that it was navigating, it wasn't the putting in of the content, it was navigating to put it where it needed to go. That was so annoying. >> That's it.

We count the number of clicks and the number of pages uh that have uh and trying [clears throat] to reduce that at the time. >> Yeah, that's the pages is where you're rolling through stuff, right? To find, yeah. >> You you click something, another page. You click another a button, another page. And I'm trying to reduce those into getting information into as smaller places as possible. >> Yeah, it's come a long way, hasn't it? I mean building apps, which I've done a bit of and and just, you know, the whole science and art of building nice elegant user interfaces has come enormous enormous way. But now it's taking a whole new direction, right? >> That's it.

And even even 2026 is the progression of the past 12 months. I mean the the numbers that I look at on that were I was excited in January this year when I could put a requirement for a a user interface to my one of my development team. And something that would have taken 80 hours came down to eight hours in January this year. Um, where we are now though is it actually now is we're now down to about two hours and it's something that I'm doing within two hours on another screen on my desk. >> So tell me more about this journey. I want to know some more.

I know you can't tell me customer details, but tell me more about this journey from 80 hours down to two and and so on because this is really exciting. And just to sort of bring everyone up to sort of what we're we're drilling in on on onto here is is we're talking about the workflows that happen within businesses. >> Mm. >> And um it's undertaking a huge revolution. And I we've done it ourselves. We took a uh uh 150 hours and reduced it to a couple of days, you know, 16 hours um so that the business was able to save $5, 000 and that was using AI and uh some software and we developed this piece for them.

Um, I'm seeing this happening across the board, this this massive speed up um and it's changing everything. So tell me more about this one. >> Well, I think if we we can plot it on we can plot it across a timeline. So as I mentioned 2013 14 I'd built 40 pieces of business logic in 12 months. Um, by 2020 we have a customer success story where that same piece of process kind of process improvement and logic we rolled out um we moved a a customer from a another CRM into Salesforce. And we by using business process and logic we five X'd their efficiency. So the same size team was able to have the five times the amount of cases and tickets and processes that came in. >> Mm.

>> Um, so the growth that was there 20 this year one of our customer success stories um is with Agent Force. And one agent that we in one piece of business logic, um the feedback was revolutionized the team. The team is now able to take on more work. They're not chasing um bits of information because the AI is able to process that. And it did and it required AI to actually do it. We did a proof of concept with um NA10 using the data to find the piece of information within an email thread and pin it back to the CRM. There was a purchase order number. Um that was the so we were using NA10 at that time and that was the proof of concept.

Then on passing looking at the security side of that for that customer and their security requirements, NA10 was working and it was great. It's a great tool. But it didn't their their security team it was outside of their their comfort zone where they wanted to be in their policies. So we were able to move it into Agent Force. >> So so just to pick up on that Agent Force, is that part of the Salesforce suite? >> Yes. Yeah. So Agent Force um became AI Force and now it's SI Force thanks to the change in the past four weeks. Um the that is the agentic capabilities within Salesforce. It started with generative but now it's agentic and can take actions. >> What does SI stand for?

>> Uh super intelligence. >> Okay. >> So yeah it's the uh with the meetings in uh Washington in the past few weeks the term's now being used. So we're uh SpaceX SI not SpaceX AI. Uh the terms coming in across the team. >> So there's these terms I'm going to get back to this NA10 in a second but there are these terms uh that people might not be aware of them. We've got artificial intelligence >> Yeah. >> which is what we're using now. And then we've got general artificial intelligence which people say is kind of replicating human intelligence. But then we've got super artificial intelligence which is way beyond what a human can do, which is certainly in the future. It's not here now.

And neither is generalized AI. But you've you've got this kind of word here, super intelligence. >> Yeah. So it's it's it's not mine. It's come from the US from from Donald Trump with the founders of Anthropic and OpenAI and all the foundation leaders and Elon's bringing it into his SpaceX term. And I think it's I mean, it is I mean, coming from the world of Salesforce, product names change constantly. So if anyone's worked in Salesforce, we're used to names of the same things changing and evolving as the product evolves. >> Mhm. >> I think it's it's it's an interesting one to to bring in now and kind of iron out the difference between AI and AGI and ASI.

So we had artificial intelligence becoming artificial general intelligence, the debate of how are we there and [clears throat] and where we are and then into the future where people are thinking of the the boomers and the doomers around artificial super intelligence of where do we go and where will we be in 10 to 15 years? It's it's a different it's a very difficult thought process to have of okay, we have this tool. It's available to us. It's it's it makes a a difference on processes. It it can revolutionize the work that people are doing. >> Mhm. >> And focusing on that and the short-term goals into and differentiating I think from are we in are we looking at Star Trek or Star Wars?

I mean, they are the super they are the big decisions of where we're going as SpaceX gets to launching millions of tons of compute into space and things that is >> Mhm. >> they they they're the things that I I look at uh outside of work. That's the >> me too. I mean, it is interesting to watch what's happening in the industry um overall. Let's come back to your NLM proof of concept because it's really um nicely grounded. So you stood up a proof of concept or a pilot based on NLM relatively easy to build and do, but you can't turn that into production. You can't use it on a daily basis because there were security issues. >> There was. >> you coded it up into your um SI Force. Is that right?

>> Yes. The um so I mean NLM we use it a lot and it's and it there is you know different companies have different security policies and different patches and >> Mm. >> it can be self-hosted on own servers. Um so those um those security concerns 100% don't apply to everyone. They are a a business's choice and we work within >> I was going to ask you about this. This is a whole complete can of worms. We'll go back into security >> Yeah. Security is is a totally different different topic. So but also I think it's the evolution of Agent Force at the time with the customer. When we first started the project, the cost of AI was high and it was preventative. >> Mm.

>> We so but by the time we got through um the journey, the cost dropped that after we proved the concept I uh Agent Force became affordable. >> So when was so when when when would you going back to to saying it was expensive? When did you experience that? >> Um I think when Agent Force I mean I'm trying to remember the year now uh would have been 2022 23 >> right, so that's kind of when ChatGPT really kind of >> That's it. Got going I would say got going cuz it was out earlier than that, yeah, 2022, yeah. >> Yeah. >> Um, but by 20 I mean, I think this yeah, this project started in 20 early 25. Um, so but by early 2026 the cost of had had worked itself out. >> Yeah.

>> we were able to meet the requirement within the security bounds and that's the end the Salesforce trust layer, the enterprise trust everything is controlled and it is secure. So it it that's the company's preference. >> That's a That's probe on that a little bit because who makes that decision and what are the factors that they if you know Matthew because it's often to me it's it's it's not straightforward with AI. I mean we're seeing people downloading, you know, health information and putting it into Claude and doing all sorts of things and I'm scratching my head going, I'm not sure you should be doing that.

>> >> Um, yeah, and I think that so there's different choices there for enterprise and small to medium enterprises and consumer use. >> Mm. >> The we are sharing a lot of information with the foundations. The you know the use of Claude is is is has exploded. I mean it's flattening out now, but and then when they introduced Fable their their model Fable they are looking they are they started you know they are analyzing the data that comes through to see what what should be allowed and that data is actually held for a period of time. >> Mm. >> So there are distinct items where we are sharing at the moment I would say without considering the implications enough a lot of information.

Salesforce have got a great setup with their trust layer that that it the the trust layer there basically anonymizes the data before it goes to an LLM and brings back. So we've that's that's a you know that great a great Salesforce have. But we're we're a an Anthropic partner, we're an open AI partner, and we're a partner with Reset Data as well. And the reason for the last one is that that's bringing in local AI. So we we I was looking at it very early, well, mid this year around okay, how much how long can we continue to be sharing IP and putting it at risk? Data sovereignty.

So we have a lot of financial services customers where we can't but they won't be able to use AI in the same extent. But we're we're we're able now to use local AI that's hosted in Australia in Sydney where the data doesn't leave the country. So we're able to do we're able to to use that within the right environments and make the decisions around what should I do with my data and and how should it run. And that's going to change a lot. The applications are are very different because the harnesses of Claude and the harness of Codex and GPT desktop as it is now, the they they're out of the box.

That's where you can sign up for a subscription, put it on your desktop and start getting the benefits of AI. I think the longer term we need to look at how we use those those models and what's appropriate. >> Yeah, I mean, the foundation the foundations, I mean the foundation models is is coming from ChatGPT Claude for anyone doesn't know that term. The foundation model means you can build lots of apps on top of a very powerful model. Have to re you know, back in the day you have to build specific apps, specific models for specific applications. So is there any guidance coming from the foundations or any limitations coming from of you're you saying that?

>> I don't think there's guidance and I'm I'm not sure there are deliberate limitations. I think there's businesses need to look at redundancy um >> Mhm. Mhm. >> because mean I've had the experience myself with one of the uh those products that we were working on on a project. It's part of our core features.

I'd started to look at okay our our own local um AI service which is wasn't up and running and the day before I started looking at the local the ability to buy credits on one of those providers had a bug and I couldn't so all my my team were messing they run out of credits uh and tokens they messaged me saying we need more and because there was a bug in the actual purchase journey I couldn't [snorts] buy more so we stopped. Now there's no redundancy available at that point of if your business processes and you can build incredible business processes using these tools from the frontiers >> Mhm.

>> but if they if the service is interrupted how much of your business is reliant and can stop on that. Having a redundancy I think is going to is is something that needs to be there. Controlling where your data goes needs to be there. >> Mhm. >> the ability to flip the switch as we can now and run local AI um and that's for clients as well who don't want us using they want us using the power of AI so we can deliver faster better quicker services um they want it on sure and we can give that security now. >> Yeah. >> We're not tied into those harnesses we're using Open Claw as our company brain. >> Yes.

>> so we're we're running that which can default to local when needed so we've got 100% residency for that. >> Nice nice. Yeah I'm going down the path of you know how to give some smaller companies some recommendations and there is this um you know you we talked about N8N and standing up a proof of concept. I know of people and I've done it myself stood it up very quickly on a weekend build an app in the morning right Only hours, I think it was. I thought, well, that's fun. That solved the problem. Um, but it's only a prototype, right? It's only it's only a proof of concept. Um, there's a long tail to this. The value seems to be there.

You come up very, very quickly, the value curve is an illusion that you're at the top, and there's this long tail where you have to kind of finish off everything. You have to get your security, redundancy, you have to control the data, and then you have to in some cases code it across onto a better platform. Um, when I'm talking to my clients, that's often a difficult conversation cuz they see it there and they go, I just want to use it. It's isn't it working? >> It's Yeah, I think that everything we build tends to be is it's being built with scalable.

I mean, NA10 is a production-ready solution that we're using in multiple integrations and we're able to put it So that piece is there, but we're we're able now when a customer comes to us with any system because we we do Salesforce, but we also build custom applications at the same time. >> Mm. >> Right? We're able to spin up the start of a prototype within a few hours so we can actually see and experience and understand the requirements at the same time and then evolve it over time without It It's part of the the lifecycle of the project. >> Mm. Mm. >> And I don't We We We put less in the bin now than we would have done in the past. And should we do it now?

Because in the past it would have been cost prohibited. >> No. >> So that journey I think is is is It Yeah, it's an exciting journey to be on. >> Yeah. Well, it [snorts] is. We've really gone down some really well-grounded topics, Matthew. It's been really interesting because workflows is where it's at. I mean, that word is coming up all the time. People talk about it in their businesses at at the top, you know, the top executives are talking about it, right down to people going, well, I work for myself, how can I improve my workflow? How can I get more out of it myself? It's just a word that's really taken off. And AI is doing it. It really is doing it.

Now, it's having this very, very tangible impact. And I tell me if you agree with me, but somebody said to me that 20 years ago, we were given all this software and we said it's going to save you money, give you a return on investment. And it really didn't a lot of the time. And I'm thinking of the old CRMs, right? Um, collected a lot of data. Now, I think it's really clear that AI is giving ROI because if you're knocking your hours down from, as you said in your example, was it 80 down to two, I mean, that that's money that had to be spent on labor. >> >> It's yeah, um, it's absolutely changing the concepts.

And I my my thoughts are that industry isn't keeping up in that thought process and they're they're thinking about, okay, what is the AI today and what how do I apply that into my business today rather than actually and I don't think that's going to give us the ability to move fast enough to actually re-engineer. And the the thought process has to be reimagining what does that work look like in 18 months time? >> Mm. >> The the capabilities of AI in the past 12 months has exploded. The I I was heard the stats the other day that even in robotic machinery, the cost of AI for a robot to do um, to do a task has fallen from I think it was $2. 36 to 36 cents in a matter of 12 weeks. >> Wow.

>> So, and that was using the new Gemma Argan from Google um, and I'm, think on that. So, that, that's really stepped up in the past few days. But the cost is, the cost of that intelligence has halving every three weeks. >> Wow. >> blows Moore's law out completely. So >> Yeah. So if anyone doesn't know what Moore's law is Now correct me if I get this right. I think it was, it was the feature size was, or the capability of the same size chip would double every 18 months. Did I get that right? >> Yeah. I think it's 18 months, yeah. It's and um I question myself when I, I'd look at was it 18 months or was it two years? It's irrelevant.

It's we're now you know halving, doubling in every three weeks. >> Doubling capability in every three weeks effectively what you said there. Wow. >> in in cost in in the cost of it. And so if the cost of that it carries on it's come, it is coming down to zero. The where and the robotics are on the manufacturing line right now to keep up with it. >> Mm. >> So the change that we've got to look at and prepare for now isn't what we have today. It's definitely what what what does it look like in 18 months time and 12 to 18 months.

And it's reimagining that process when I look at a a business and think okay what the processes that we do day to day and we've done this ourselves um can be fully automated using AI. So we um our processes we we're using uh Claude and [clears throat] Codex to we have a a conversation with a customer. Everything's recorded. That becomes a business requirements document uh or a pains pains and goals. And every step gets produced into fully documented um human reviewed processes and scopes. The the thoroughness that we're able to do that within the time frames mean that the service we're able to offer is >> Mm.

>> um I think that 80% of the tasks that we do can be automated into that process, which means the same people can handle multiple bits of work, which is what our customers are are finding as well. >> Mm. >> Do more. Um. >> It's funny how that 80% comes up again and again. That was a metric we used in one of our projects, and it's the 80/20 rule, I think. We we're talking about the famous 80/20 rule which says something like 80% of the benefit comes from 20% of your >> Yeah. >> efforts, I think is the term. Yeah. >> That's it.

So if if everyone becomes five times more efficient because they're they're doing the 20%, how do we give them a user interface, a the the human experience to be able to keep up with that level of work? We need to be able to um I mean AI um when when we ask AI to do documentation and we don't give it any guidelines, it comes out with huge volumes of information, which is completely >> looks nice until you read it. >> >> Well, yeah. I mean it looks great. How many I've only got 30 pages to read, but um it yeah, if you read it actually it doesn't you can't understand it.

So you know, we have to control it, but in the same way we need to present the CRM needs to present the information in the in a way that can manage it. So it's how does work look. So if we were in 20 10 years ago thinking, okay, well how many clicks and how many pages, it's now five times that, so we're back at the same page. We're back at the same point that to do that much work, the human has to be able to keep up. Um and and we need to be able to think about those interfaces and change those as we go. >> Yeah.

>> So a lot of the the lot of the pieces that I've been doing and building within some of our I'm software that we're we're now producing is about the agent being able to control the screen. So uh is the future just talking to a chatbot? Well, if there's that amount of text and we're reading it, even if we're speaking to it, we want to be able to visualize that information. So we need the screen to be able to flex as it goes. So can it jump from a lead to an opportunity? >> Mhm. >> By just the conversation to show so the effectively your agents and AI is showing you the information relevant to your conversation so you can help visualize and context it. >> Mhm. Mhm.

>> It gives the gives us the ability that as a user I can go in and change a field or I can let the AI do it's both. >> Yeah. >> So you've got both options of and then it's you can sit on your phone and talk to it and knowing that it's being updated >> Mhm. >> or you can sit at your desk and see it being updated. >> Mhm. >> We have the flexibility and then I don't know, 36 months, is that actually a just in the headset in the glasses? I mean >> Right. >> with all of the the visual technology um side of being able to see you won't need a computer screen. You can be able to to your AI and it can show you the information uh into your direct >> Amazing. We We got too much to go into today.

So running out of time, but I wanted to ask you uh one one my second last question, which is are you seeing some downsides to this? It seems like it's really going to help businesses with your product that you're working with and the workflow changes. Have we seen some downsides? >> Um I think there's I mean there's we can't deny that there is going to be there is a shift in the workforce, but it's about retooling and re uh working the the the teams that we've got and the best can.

If you the um I think it was IKEA that took the um a huge cost saving out of their customer service team and but rather than moving them out of business, they retrain them to become stylist consultants um and actually reappropriate them into uh possibly probably certainly more interesting work. So I think the opportunities that we're going to generate um are there. Finding them can take time and >> That's a good point.

I heard that a couple of times over the last few days about um at the company level, AI gives us not just the opportunity to cut costs, which might be cutting staff, but also reinventing new um differentiating products and services that would make us more competitive to to to our competition, make us stand out, do something different, do something more, do something new that hasn't been done before. Stylists versus I guess they were packing staff. I'm not sure what you said, but >> When when we ask AI to do documentation and we don't give it any guidelines, it comes out with huge volumes of information, which is completely >> looks nice until you read it. >> >> Well, yeah.

I mean evolve quickly looking ahead and and planning and finding those opportunities. >> Mm. >> That's where I think it's going to be and 2027 is going to be the year for that. >> Yeah. Tons and tons of innovation. So Matthew, you got a lot here. We're going to have you back on, I think, to do some more. Do you have any questions or comments before we we start to wrap up? >> Um no. I can talk on too many different subjects. The and I think the world but I mean comments would be yeah, the the opportunities and that we're going to that we're seeing and can be found are are going to continue to evolve because there's so much change that that is going to be there.

Um I made the decision very early on to embrace it and to make sure I was ahead. Um it took me, you know, a good amount of learning to go from okay, what's AI within Salesforce and Agent Force to actually >> Mm. >> what is AI in general? Because the Salesforce released headless 360 um, as the concept that it doesn't you don't have to work in Salesforce. So where am I working? What are the systems am I um, [clears throat] interacting with and how do I create efficiencies in this and so there's going to be we're at the beginning. >> Mm, yeah. We are at the beginning of a huge transformation. So one final question for you today, Matthew. AI, is it hype or help?

>> [clears throat] >> Uh, well, there's 100% there's a lot of hype and that's I think we've got to split it because the hype is we don't know what the future is going to be. >> Mm mm. >> Where we are but where I absolutely believe in is that it's a help. What we have now right now can help us with so many of the um, day-to-day um, business challenges, but actually stresses that we have of of how we operate as human beings and you know >> mm. >> our our roles have been data entry um, and use the CRM. A bad CRM is used as a reporting tool. A good CRM is used as a sales tool or a customer experience tool. And there's distinct differences.

And AI helps us do that by taking away the pieces that we don't like doing. >> Yeah, nicely said, Matthew, man. Now, can you stick around after after we finish up today? But I want to thank you for coming on. And like I said, I think we should come back because you're right in the sweet spot where you're rolling out tools and implementations in workflows uh, for businesses. And that's what we like to talk about here on this podcast. But for today, thanks for hanging around, Matthew. Really appreciate the the conversation. >> Thank you. It's been a it's been a pleasure. >> Thank you. Well, >> that was great talking to Matthew.

Very grounded in workflows and practical implementations of AI, which are having real impact on businesses. The product that we talked most about was Salesforce and a feature within that called Agentforce, which is Salesforce's AI-enabled capabilities, which touches sales, service, and operations, and how it's revolutionizing the workflows. So some of the metrics we talked about were the the the number of people required to do the same task, five to one ratio. In other words, one person is doing the work of five, or down from down to 20%. So massive, massive labor costs. We talked about hours. One project used to be 80 hours, and then down to 12, and then down to two for the same task.

And finally, we talked about costs, the costs of actually driving the Foundations models. For doing particular tasks, costs are coming down by 50% every three weeks. Now we we correlated that with Moore's Law, which was 18 months. This is such a rapid reduction in costs. If you want to know more about Matthew and his business, please follow the links in the description. If you want to know more about Skillion, please do the same. And we'd really like it if you could subscribe to this channel, hit the like button, and share it out. It would really help us get this channel out there and get us more support. The algorithms really need those interactions from you. So thanks again.

I hope that helps, and thanks for watching.