How To Run An AI Transformation Project
This week, our most recent episode of AI Moment went live, and it’s a big inside look at how to run a real AI transformation, not just a series of tools.
Listen on Apple Podcasts, Spotify, or YouTube.
The Breakdown
To run an AI transformation successfully, you must treat this project as a business transformation rather than a simple tool rollout.
I highlight a 6-step framework to move from experimentation to scaled implementation:
1. Map Your Workflows (Timestamp 4:24 - 6:35):
Get your team together to document how work actually gets done. Identify "unknown knowns" the informal, undocumented processes that keep the company running alongside your formal processes. Automating with AI tools sounds great, but there is no quicker way to find out what’s broken than automating with AI.
2. Identify Opportunities (Timestamp 7:41 - 14:23):
Use an audit to categorise tasks into four buckets:
Accelerate: Tasks that work but take too long (e.g. drafting proposals).
Optimise: Processes with bottlenecks (e.g. lead handovers and handbacks).
Cure/Fix: Areas with inconsistent quality (e.g. customer service responses).
Kill: Manual, soul-crushing tasks that can be fully automated or are just wasting time and resources (e.g., data entry).
3. Plan Training and Risks (Timestamp 15:10 - 16:15):
Before scaling, run workshops and hackathons to build team capability. Critically, map out risks (security, accuracy, judgment) before calculating potential benefits or revenue gains.
4. Assign Ownership (Timestamp 16:35 - 20:47):
Establish a cross-functional SWAT team to lead the effort. This prevents departmental silos and ensures accountability, as AI integration requires shared knowledge across functions like sales, marketing, and finance.
5. Apply Governance (Timestamp 21:31 - 23:22):
Implement a risk framework to guide human judgment. High-risk decisions such as external communication or committing funds, always require senior expert oversight.
6. Execute in Sprints (Timestamp 23:36 - 24:43):
Run your project in 30, 60, 90, and 180-day segments. This allows you to secure quick wins, pilot effectively, and eventually achieve scalable transformation without overwhelming the organisation.
My Golden Rule:
Don't start by chasing tools or AI agents. Start by fixing your processes, centralising your company context, and ensuring senior leadership is actively involved in the transition.
Do you need help to roll out AI in your business?
Get in touch ai@dannydenhard.com or hit the button to book straight in for a call
Listen below on how we break down AI transformation projects and how we challenge each other on the order, especially around data and understanding that AI workshops and AI hackathons need to be included.
Full podcast transcript
[00:00:00] Jonathan: You're listening to the AI Moment with me, Jonathan Wagstaffe
[00:00:04] Danny Denhard: And me, Danny Denhard
[00:00:06] Jonathan: And this week we're going to look at AI transformation, and we're going to look at that because one of the most common conversations that both Dan and I have had with businesses in the last few months is lots of people playing and experimenting with AI, but few businesses making the real transition to starting to embed it in the organization, starting to get real business benefit from it, starting to really use it as a core part of what they do. So I, you know, I think setup for this is AI transformation is actually a business transformation that happens to involve AI, and the mistake we're both seeing at the moment is that are treating it as a tool rollout. So they're buying licenses, they're, they're playing with it a little bit, they're encouraging their people to experiment, and then they're wondering why nothing really changes. So that approach of like, 'Here are some tools, go use them,' is not working for people. And maybe a stronger approach is to say, 'Here's how [00:01:00] this business creates value. know, let's start to analyze the workflows and find the bottlenecks, the knowledge gaps, repeated tasks, where are the decision points, and then let's work out where AI can help.' And so today we're gonna unpack that and maybe try and give people a bit of a framework or a guideline for how you might go abro- about going from experimentation through to piloting through to scaling and implementation on a, on a bigger scale for a business.
[00:01:26] Why AI Efforts Fail
[00:01:26] Jonathan: So, you know, to start that process, I, I mean, I've got a few things down that, that, that I'm seeing, I think.
[00:01:32] Jonathan: There's... The reasons that people's experiments or projects are failing at the moment, I think there's, there's five or six in a list maybe. One is that, uh, I think it becomes too tool-led, so people get very, very focused on should I use Claude or ChatGPT or Gemini or Copilot, and they're really all over that decision without thinking about the business outcomes and some of the broader things.
[00:01:53] Jonathan: That's one. I think I'm seeing, um, departmental initiatives that aren't [00:02:00] cross-functional, so maybe marketing's doing some stuff with content, sales is doing some stuff with email automation, finance is doing some analysis, but nobody's swapping stories or getting to work together. Um, I'm seeing quite a lot of experimentation where people are experimenting a lot but not trying to implement, or they're ignoring the existing processes in the business and so the experiments are irrelevant. Um, and then, and then the last couple of things I'd add, and feel free to add to the list. The last couple of things I'd say, quite often I'm seeing it lacking in ownership. So I've spoken to a number of businesses where there's sort of different departments playing with it, but nobody's owning it, and certainly not at the senior level.
[00:02:35] Jonathan: Nobody on the board has got at least part of their job to be responsible for the IT implementa- or AI implementation, I should say. And, and also people are also not using judgment with it, so they're either trusting everything it tells them or they're not trusting anything it tells them, and that, that balance, people are either trusting it or rejecting it too quickly. So- AI transformation isn't failing because people can't use the tools, it's [00:03:00] failing because the organization hasn't decided beforehand what better will look like, what better work will be. W- what do you think? What have I missed there, Dan?
[00:03:09] Danny Denhard: I don't think you've missed anything. I would say that you've probably echoed what 70, 75% of people have said out loud and probably said in frustration within their organization, whether they're, small SME- SMBs, SMEs, or they've gone up to, you know, like really big listed companies.
[00:03:26] Danny Denhard: They're all, they're all struggling to, really try and create an AI transformation project. And like you said right at the start, many people haven't considered it as that, and it's just been a directive thrown into the ether or a goal that's been squeezed on top of other company goals, whether it's an OKR or part of your EOS system.
[00:03:47] Danny Denhard: I would say you've said it a number of times and you've reiterated it just now. There isn't a leader who's leading from the front, so it's very difficult to have that executive presence, that, that importance provided [00:04:00] onto it, or IT have to pick it up because it's always been given to IT it's a license issue or it's a token issue, so it's something they've done.
[00:04:11] Danny Denhard: When people are doing it well, where, where do you think people-- how do you think people have landed, Jonathan? What things have they done well or what makes them successful when, when it does go well?
[00:04:22] Map Your Workflows And Share Your Knowledge
[00:04:22] Jonathan: Well, I think that takes us nicely onto the start point I think, Dan, that we both agree on, which is the start point is really to map out the existing workflows and the existing work that happens in the business. So, so take whichever department you're in, take your processes and, and put them on a whiteboard, get people in a room and write them down and look at the cross-functional processes, looking at the processes within departments. But start to write down, um, how the workflows in your business work, how the processes work, where the information gets handed off, where the decisions get made, where the communications happen. Um, and as we've talked about before, you know, uh, [00:05:00] sometimes that's a useful exercise anyway because actually you'll find that even within a department, people don't quite agree exactly how we do things now, and all we're doing at the moment is writing down supposedly how we do things now. But you might well find there's significant disagreement around that. I think that's the start point. What, what are the sorts of things, Dan, that you'd wanna cover in there in a session like that?
[00:05:20] Danny Denhard: Yeah. In part one of what I would say my six-step process as you said, list and map out your issues. So your known knowns, uh, everyone's sort of familiar with that. Your unknown knowns, which are, the things that actually happen and are understood or processes but fail to be like formally written down or recognized.
[00:05:39] Danny Denhard: So they're...
[00:05:40] Jonathan: Mm-hmm.
[00:05:41] Danny Denhard: That's often the ones that I find are, are the most contentious or the things that people just are completely unaware of but are really important. So that one person you go to to fix something that's never written down, but everyone knows it in their head. So there's numerous different examples of that.
[00:05:57] Danny Denhard: I like to put down what works, [00:06:00] um, and then what... Then you know what becomes scalable, what's broken, and what breaks often, what fails and where, so then you can understand how to like fix it or apply AI, et cetera. Very similar to what you said.
[00:06:13] Danny Denhard: And then once you've got that down, whether it's on a whiteboard or more formally, you can create your company context document, as we've talked around before.
[00:06:21] Danny Denhard: That's a way that you can formalize, share it around, get people to input into it, hopefully not too many people, get sign-off, and then you can start answering some of the, some of the problems and some of the scale and, take on some of the scalable things that you've highlighted. So that takes us on to process two.
[00:06:37] Jonathan: Couple of things to add to what you said there. I think one is to identify... The, the low-hanging fruit for me is where the real pain points are. So that process you just talked about, if we can, if we can find some real pain points in there where we're, we're wasting time, we're repeating things, people find it an absolute pain to do, they're the places where you'll get big traction very quickly. And I think one of the key questions in there is that-- is to [00:07:00] ask the question: what do people know individually that the business hasn't captured? That's a really important thing to do in that point. Most businesses actually know far more than they realize at any given point. The problem is that the knowledge is trapped in people's heads, or it's in documents on some old server that nobody's looked at in five years, or it's in a Slack thread.
[00:07:14] Jonathan: It's in the CRM, and the CRM's not reliable. You know, there's all-- sales decks. There's all sorts of places, as you know, that the stuff... You know, even just habits that people have. People just intrinsic- "Well, this is how we do it around here," but nobody's written it down. Capturing all that, you know, the AI then can really become more useful when we surface all that knowledge and we structure it and we connect it. leads us on to, okay, what do we do about that? As you say, the next part of the process.
[00:07:41] Opportunity Map Accelerate Optimize
[00:07:41] Danny Denhard: number two I call it answer mode. So it's like answer part of the process. So what can be accelerated? So build faster or built faster, less reliance on it. Then what can be optimized, and so where the bottlenecks can be removed. What can be fixed or improved? I think that's one of the core areas people are going too big [00:08:00] as opposed to fixing the, the common issues and the, and those common threads.
[00:08:04] Danny Denhard: And then what can be removed? I'm a huge believer in lean methodology and going back to first principles. That is something that so many people forget is such a critical part of any transformation, s- more specifically AI, and that's where most people can win,
[00:08:19] Jonathan: So I think if you take each of those in turn, what can be accelerators? I think tasks that we can see when we do the mapping process. Those tasks work, but they take too long, and we think there's a real opportunity to accelerate them, as you say. of example... I mean, I'd be interested. For me, it's like things like drafting sales proposals might be in that. Um, things like creating training materials might be in that. Were any other thoughts on things that might fit into there?
[00:08:44] Danny Denhard: Yeah, there's a lot around reforecasting, so anyone that goes through any budget cycles, that's one of the most frustrating things that you have to do as a, as a business leader. So that's something that can be probably built more quickly. It doesn't have to have 17 variations of the spreadsheet. [00:09:00] People don't get in trouble for using color coding or not. There are things that can be optimized.
[00:09:04] Danny Denhard: There's a lot within marketing and product that ca- that can be accelerated, especially around process, prototyping, et cetera, any sort of
[00:09:14] Danny Denhard: campaign management optimization. These are all key parts that can be accelerated and identified, and then, you know, you can plug into Claude or, or your preferred LLM
[00:09:24] Jonathan: And actually thinking while you're talking there, one of the ones that I do across all the time is, you get Zoom to transcribe the meeting you have, and it gives you a, a summary of the meeting and some actions straight out the bat, and you just copy that into your CRM. And that's, that takes a minute at the end of the call, which would've been 15, 20 minutes of typing stuff in in the old days. So if you move on to your point two, what could be optimized? We're really talking here about processes that basically do work, but there are bottlenecks in them or there are inefficiencies in them, right? So, uh, immediately I'm thinking about things like lead handover from marketing to sales is one that in our world is a very ob- Any others you can think about there?
[00:09:57] Jonathan: I think, I don't know, customer onboarding maybe. What, what sorts of [00:10:00] things would you,
[00:10:01] Danny Denhard: Yeah.
[00:10:02] Jonathan: see that gets you straight away?
[00:10:03] Danny Denhard: Customer onboarding, employee onboarding, uh, anything that has been highlighted in, you know, customer service or customer support, and that can be transferred over. Um, how can that be synced more, you know, more quickly and optimized? How can we remove that, that delay? So as you said, you know, whether it goes from MQL to SQL, for instance, or PQL, and then you need the sales led motion on top.
[00:10:28] Danny Denhard: All of those are often slow. There's no reminder set. The quality scores aren't set properly or, the boundaries or the guidelines aren't set. When the sales team say it's not high quality enough, we're not gonna call them. All these things can be optimized and you can remove a lot of those bottlenecks.
[00:10:44] Danny Denhard: So some-- I know some businesses that will reduce the amount of contact that they have with customers. But actually, then once you have that delay, the customer forgets mostly about you if you're not the brand or the brand that you've reached out to. So these are all [00:11:00] bottlenecks that can be optimized and improved via CRM, via, a number of different out-outlets and departments
[00:11:07] Jonathan: And again, thinking while you were talking there, uh, one, another one that struck me in the last year is HubSpot. So, you know, creating a marketing or sales report in HubSpot would've been 20 minutes of pointing, clicking, and dragging things around, and now you can do it with a, within the HubSpot AI. You just say, "I'd like a report that gives me the monthly sales figures," da, da, da, da, da, and you get this thing in minutes.
[00:11:26] Jonathan: So maybe that's another example.
[00:11:28] Danny Denhard: Like you were saying, reporting is the other one, right? It, reporting is the one thing that everyone can do, and there's a lot of ways you can automate in dashboards, but actually it's the analysis part that's the most important. So, so what? What are we gonna do next? What's the most important actions we're gonna take?
[00:11:44] Danny Denhard: They're the types of things that can be optimized, especially if you use voice, uh, and you can interact with a dashboard. So moving from report to analysis to actions is, is one of those things that I think are gonna be vitally important
[00:11:57] Jonathan: Agree. Would agree. And then the third, your third point, what can be [00:12:00] fixed or improved? So this is where quality's inconsistent in particular. So, you know, to your point a minute ago, sales lead follow-up, quality of customer engagement and start the sales process. Just about everywhere I worked until maybe a couple of years ago, that was really down to the individual quality of the salesperson. A- and even with all the training in the world, there was a great inconsistency there. Proposals were inconsistent. Customer service respon- you know, there's a lot of those sorts of things. And is that the sort of thing you're thinking about when you're looking at that, Dan?
[00:12:28] Danny Denhard: Definitely. There are also ways where you report if you, if you're owned or you've had significant investment from a PE firm. Very often it's the way to understand how to report to them and what the most important updates are. How can you remove some of that, uh, disconnection and actually take on board what they've said, work with AI or work with your PE partners or, you know, your VCs to work out what they really need, and then you can fix it or you can improve what, based on their feedback or based on some of the signals you're getting [00:13:00] from the noise within your business.
[00:13:01] Danny Denhard: So they're exactly, they're, they're the sort of areas I would say are key to improve
[00:13:07] Jonathan: Then your last point was what can be removed? So this is work that shouldn't exist, right? Or that could be automated. So I get-- I'm guessing things like data entry could be automated. Manual reporting as we just talked about. There's... What, what sorts of things are we talking about there?
[00:13:21] Danny Denhard: It-- Your example's the best one, actually. I think when you're updating HubSpot records or you're up-you're updating a number of different processes. So one thing you said, and that has stuck with me ever since, is, you know, salespeople or marketing people hate updating CRM systems.
[00:13:37] Danny Denhard: So actually, a lot of that can be done automatically or can be done, um, done async via, you know, different methodologies now. So it could be fired off from the CRM or from Zoom or from Gemini, et cetera. Um, so if you've got Google Meets, for instance, you can, it can update a record or it can update automatically when you've contacted that customer or they've contacted you.
[00:13:57] Danny Denhard: So it's st-sunk across all of the, the [00:14:00] process, so you can remove a number of those steps. Another one is updating documentation. If-- when it's done automatically or automated, that can be completely removed so that they're key areas.
[00:14:09] Danny Denhard: Also, one thing that I love is removing some of the automate-automated emails, so it's updating CRM flows, et cetera.
[00:14:16] Danny Denhard: A lot of that can be removed, but until you audit it, you don't know what you can, what you can keep, kill, cure, basically
[00:14:23] Jonathan: And I think you too, Greg, 'cause I mean, your point there about find that when I talk about the salespeople using the CRM as an example, when you map your workflows and then look for the big pain points, whenever I talk about that as a pain point, there's recogni- every single room I've said that in, there's recognition in the room that that's a massive pain. And the other one that you've just alluded to, you know, when I ran an IT business, getting the technical guys to document what they did was such a pain, right? And if that could be automated, that, that, that the AI could just look at the configs of the systems and record them, that would... That would fix a huge problem in the IT industry, so we can see that. so we've done that section, Dan. [00:15:00] We've looked at, you know, what, what's the opportunity map, what's the accelerate, optimize, fix, remove, elements. So where, where do we go from there? O- once we've done that, what, where do we go from there?
[00:15:10] Danny Denhard: Yep.
[00:15:11] Hackathons Owners And Risk
[00:15:13] Danny Denhard: Step three is map- mapping out or map out. I would say you should start with training, workshops, hackathons. So we ... Funny we didn't lead with that, but I think once you get all this down, then you understand where people need training, when there can be workshops, where there can be hackathons. So that's at team level, department level, unit, or company level, and that's really important.
[00:15:31] Danny Denhard: I think if you don't map that out and then understand how to drive that forward, you're, you're only inviting more siloed activities as opposed to a company-wide, and then you can connect it to really specific goals.
[00:15:42] Danny Denhard: Then I'd map out the risks. So before I say risks and benefits, so before you do the benefits, I really would, suggest that you map out the risks first, 'cause then you can understand how much work, effort, and how much training needs to go into it, and then map out the benefits, and then work [00:16:00] out the revenue, whether that's a loss or a gain, and then what, and what time period that is, Because you can say all these big benefits, but actually if you don't map out any of the revenue, or you don't anticipate that, s- the CFO or the FD probably won't sign it off anyway.
[00:16:14] Danny Denhard: So it'll be training, workshops, hackathon, and then map out risks, benefits, and then revenue loss or gains
[00:16:23] Jonathan: Are there any prerequisites in terms of getting your data right or any of those sorts of things before you get into the hackathons and the, the deployment bit? Is there anything, anything that people need to do on top of the things we've already talked about?
[00:16:35] Danny Denhard: Yeah, I'd say that's actually, um, in mine that'd be number four, so that'd be assign team members and owners. Because one of the challenges that you've got with data is a lot of the time there's probably different owners or there's different collaborators on it. So there's probably an owner and then co-owners, and actually until you work out what training you need and until you work out some of the workshops and some of the hackathons that's needed, when you work that out and [00:17:00] you work together, then you can start working out who should be a leader, who should be in that SWAT team to, to take on that big, that big transformation project. I would say the data is important, and you could map it out in the third section, but I would say it actually rolls into the fourth for me personally.
[00:17:16] Danny Denhard: But happy to, uh, happy to let that one go into the third part if that's something that, uh, is vitally important or is in multiple different areas and needs to be centralised first.
[00:17:28] Jonathan: Yeah, no, I think there's, there's some things around, you know ... So we need to, we need to get the data right. We need to give the AI relevant context so that, you know, when, uh, around, okay, what are we trying to achieve with it, with it... As you say, "Okay, let's put some people in a room. Let's do a, let's do a lab, an AI lab or a hackathon."
[00:17:45] Jonathan: But, but I think there's some context maybe to put around that. So what do we... What does the AI need to do this particular lab project or hackathon well? And so what are we, what are ... So we're making sure we- our data's tight or our, our thinking's tight around that sort of stuff. And [00:18:00] then, and then tied into that, I think within the guidelines of those hackathons will be, you know, if necessary, build a new system.
[00:18:07] Jonathan: But even better if you can make the AI work with the current system and optimize it, rather than go and build a new AI-based system just for the hell of it. I mean, m- you know, if the process is messy, then AI might make it mess faster. But actually we, we use the AI to try and improve the current system rather than, rather than wr- write a new one in.
[00:18:26] Jonathan: So if some guidelines I think round the, round the experimentation, um, and then also, uh, obviously getting the cross-functional piece going is really important there. So, you know, what does each, what does each function need to bring to that hackathon to really make it fly so that you don't just get some generic slop out of the back end that the, the, each
[00:18:44] Jonathan: You know, sales is bringing objections and buying processes and opportunities, and customer service is bringing recurring issues, and marketing is taking those recurring issues, turning them into content and mess- ... You know, they're doing that kind of thinking where they could, they could pull off each other. And then I guess the prioritization then is, uh, once [00:19:00] you've done the hackathon or even before you do it maybe, which of the things we're gonna work on first, which are the, are the, which are the ... And that's down to your point about, you know, which opportunity, what's the business value, what's the risk, where are the wins to be had maybe around that.
[00:19:13] Jonathan: That, that might be the, um, that, that might allow us to decide where the hackathon should play and which are the use cases that we're gonna go for first possibly. I don't know, what do you think?
[00:19:21] Danny Denhard: Yeah, I don't, don't disagree. My point around more the data and context is once you start assigning people and projects, then you can start working out what the most important parts of data it is or what context that you start building for, for the AI or for the company, because a lot of this is pre-AI, right?
[00:19:43] Danny Denhard: So a lot of this is get people on the on the same bus in the J-Jim Collins sense. It's getting people to understand this is our way of doing things, getting people confident in collaborating first. Then you can start assigning owners to, to take it forward. I personally,
[00:19:58] Danny Denhard: If we were to move on to section [00:20:00] four, I always assign SWAT team. So I always assign really specific people within businesses to then go and it's like a collective of small, agile working people, very smart, who are then known to bring in different experts based on the projects or based on the initi-initiatives that you, you come up with.
[00:20:17] Danny Denhard: So you need... That SWAT team, you need a leader in it, and then you need something like a rice methodology where you improve communications, you, uh, create accountability, and then you know who's involved when, and you know how to do the communication process and the cadence.
[00:20:34] Danny Denhard: Then you can start working out the data in a context. Because if you don't have that team or those core teams on those core projects that you've decided, it's very difficult then to work out do you need to actually move data or is it that you need to put RAG on top, for instance?
[00:20:48] Jonathan: Well, and also, you know, you've talked before about human judgment in these situations, and I think when you risk managing what those AI projects might look like, human judgment bit's really important, isn't it? That, you know, you've [00:21:00] talked about that before, and it's, it's, it's really important that you get that piece right, yeah
[00:21:04] Danny Denhard: Yeah, no, if you assign people to projects and they're, they haven't had that level of training or they're not that comfortable in it, or they're way ahead of their, their immediate colleagues, it's very difficult for them not to always be that expert that everyone always goes to as opposed to that unit who are moving forward to, to take on those really deliberate projects that you've identified.
[00:21:25] Danny Denhard: That's key for me and that human judgment is always something that is, uh, should ring true for every project really
[00:21:31] Jonathan: Yeah, and I think you could, you could almost decide that's, that's on a kind of high, medium, and low-risk basis. So, you know, low-risk AI is like drafting documents, summarising documents, helping to organise data, brainstorming, quite internal stuff. know, th- then you move up to a slightly higher, which is if you're asking it to recommend or analyse or prepare things, that's a bit more of a risk 'cause that might be external. And then high risk is if you're giving it agency to decide or publish or send emails or commit money [00:22:00] or, you know, reply to customers, engage with customers, that's higher risk. And the higher the risk space that you're asking it to work in, the more human judgment and senior expert human judgment is needed about, are we gonna turn this on or not?
[00:22:12] Jonathan: Are we gonna do this or not? I think that, you know, that, that to me makes sense as a, as a, as a, as a, as a way of operating. But I think it's, it's down to each business to decide which are the high, medium, and low risk items for us and what level of, you know, individual governance do we need to put on that. If you move
[00:22:31] Danny Denhard: Oh
[00:22:32] Jonathan: to next stage, Dan, what are we, what are we talking about in the next stage?
[00:22:35] Danny Denhard: I like to say keep, cure, kill, copy. So run the audit on those projects and then start working out. So keep what you're doing, keep what's doing well, what, that can't be optimised. Cure or, optimise, depends which way you wanna say, is what can be improved, because then you can start building those projects around it.
[00:22:53] Danny Denhard: Kill is the things that, you know, like we were saying, waste time, isn't effective, um, actually if you try to [00:23:00] optimize it, it'll waste too much of your own time and that team.
[00:23:02] Danny Denhard: And then copy what other people are doing essentially. So it's a very deliberate way to go out and understand that, and then connect it to what's being worked on already and what's in roadmaps, and then understand actually how do you assign a RACI or, or a DACI to that to, to work out and break those projects down so people understand what to do, how to do it, when to do it
[00:23:23] Jonathan: I like that framework. I like that, the way you're framing about the, you know, how you take each of those things forward. that, does that relate to ... I know you've talked about a kind of now, next, future approach. How would you map what you just said onto that? What's the, what's the now, next, future bit of the, uh, of the implementation?
[00:23:37] Roadmap Phase Two And Wrap
[00:23:37] Danny Denhard: So now is essentially what you're gonna do for the next 30 days or 30 to 60 days. Whether you run it in sprints or not, that's entirely up to you. If your business is agile or, um, or understands that, then you can do it.
[00:23:50] Danny Denhard: Next would be 60 to 90 days.
[00:23:52] Danny Denhard: What's really important is if you run it in 30-day segments, then you can understand what needs to be rolled over, what is or isn't working, or what can be worked a little bit harder.
[00:23:56] Danny Denhard: And then the future is 90 to 180 days, so 90 to 180 days [00:24:00] out from there.
[00:24:01] Danny Denhard: And then the bigger projects you can start assigning, so some will take longer than others. For me, it's a case of you need these sections of time, these sprints, because otherwise it's very difficult for you to decide what to do, when to do it, how to do it, and what, uh, resources are needed.
[00:24:18] Danny Denhard: And I wouldn't... There's two ways you can do it. You can either integrate it into one roadmap, which is quite difficult, and then you have essentially have two ro- roadmaps, or you have a plan which then runs alongside it. I think it should run alongside it, and then they should connect it so the better communication, the better the optimisation can be.
[00:24:37] Danny Denhard: And then when they over cross or they connect, is then you can optimise it and bring the wider team into it as well
[00:24:44] Jonathan: I think that's a really... I, I really like that practical framework again, the 30, 60, 90, 120, 180. So, you know, if, if you're starting from scratch, next 30 days might be you form your small cross-functional AI working group. You do that workflow [00:25:00] mapping and process mapping we talked about. You maybe, you maybe capture the unknown knowns.
[00:25:04] Jonathan: You identify some of the key pain points, and you maybe t- pick two or three, um, use cases to go and do the hackathons and AI lab stuff we talked about. So you kind of do that within a month, agree some principles and guardrails, then, you know, then maybe on... in the 60 to 90-day window, you run some of those hackathons and those pilots. You start to find what success looks like. You start to train people a little more.
[00:25:28] Jonathan: You start to be able to measure maybe how well you've done with it. And then you can start to roll things out into operating practice. That 90 to 180-day window, keep improving, you know, starting to connect cross-functional systems. That would be my read. Is that, is that fair?
[00:25:42] Danny Denhard: Yeah, I agree, I agree. And that's something that I think is vitally important, is having it all mapped out, knowing it's iterative, and then understanding the goals, the metrics, the revenue figures, the previous failings, so then you can decide whether you're trying to optimize again to it.
[00:25:58] Danny Denhard: Obviously, there's that layer of human [00:26:00] judgment, as you said, that has to be included all the way through.
[00:26:03] Danny Denhard: And then you can work out, never rush towards fixing big and the biggest problems, because you're setting the worst expectations, for you to do it, especially with, with AI. Um, centralise a lot of the information and as you go, you have to centralise that and add more and more context to it.
[00:26:19] Danny Denhard: So whether that's obviously, like you said, the data, whether it's email, whether it's interviews, documents, et cetera, that remove a lot of the friction.
[00:26:28] Danny Denhard: Don't start at job cuts. Like, I know it's where a lot of people start, but that's not transformation at all. That's... All you're gonna do is, is hinder your company.
[00:26:37] Danny Denhard: Don't jump into agentic first, because all you're gonna do is you're gonna be spending so much time trying to work out what we need to automate, what needs to be, what breaks all the time, what needs to be fixed all the time.
[00:26:48] Danny Denhard: It's just gonna be start, stop, start, stop, start, stop. W-
[00:26:51] Danny Denhard: And then start with smaller business areas that have, that can scale well. So customer service, CRM flows and CRM data, [00:27:00] security, and probably sales, 'cause then you can increase and level up the quality.
[00:27:04] Danny Denhard: And don't get into that trap of start-stop because all you're gonna do is stop every time something goes, uh, doesn't go as well as you think.
[00:27:12] Danny Denhard: I'd just be really mindful of that and just know how critically cross-functional knowledge is, and just think when people go on holiday or a vacation when they go away and... you have to wait for the week, ten days, fourteen days for them to come back to get that, all that sort of information should be added in as you go, so that context is really set all the way through.
[00:27:32] Danny Denhard: And that is how I believe you should do AI transformation projects.
[00:27:36] Jonathan: Yeah, no, I completely agree. And when you stand back a bit from that, we talked about the, the big steps really. We talked about mapping your workflows, finding the pain points, you know, automating those pain points, running pilots as kinda phase one.
[00:27:50] Jonathan: Phase two then is, is the, okay, so we've automated a lot of good things here and saved some money. But phase two, and the real win as we've talked about before, is what could we do [00:28:00] that we could never do before with AI? But to do that phase two, you've got to have learned the lessons out of phase one. You've got to have looked at your business, really measured, analysed it, cleaned up your data, sorted out your workflows, understood where AI can help.
[00:28:13] Jonathan: And once you've done that apprenticeship, and you've done the bit that saves time and cost, then you get to play on the second field, which is where ... Then it's like, right, what could we now do that we've never been able to do before? How could we really accelerate this business or really 10X what we're doing?
[00:28:28] Jonathan: That's when you do play with agents. That's when you do a lot more of the experimental stuff. But you can only do that if you've done the first bit well. And back to what we said at the start, I think you've seen people messing about with agents, having not really set the groundwork properly, and that's why those things aren't working for them.
[00:28:43] Danny Denhard: The training and the co-training. So once you got the training, it's the constant training and co-training. So it's colleague to colleague, department to department. That inter-training, that interpersonal training that's im- that's frequent, that's ongoing, it could be every Friday morning, is, is gonna set [00:29:00] you apart from your competitors because they're not doing it properly.
[00:29:03] Danny Denhard: So if you can, you, you'll win
[00:29:06] Jonathan: Yeah, and th-this, this question of how do you do it is, is why we've done this episode because so many people are like, "I wanna do this, I just don't know what the framework is." So I guess if I could summarise that framework, Dan, you're really saying... We talked at the start about a transformation, not a tool rollout.
[00:29:21] Jonathan: We talked about some of the reasons they're currently failing the projects. And then really you were saying, first of all, we've gotta map the work. We've gotta know what works, what breaks, what fails, what knowledge is trapped in people's heads. We do that bit. we go into using AI to identify what could be accelerated, optimised, fixed or removed. Then I think you talked us through the idea of running some hackathons and some projects, pl- pilot projects to get cross-functional teams together to work on specific use cases, get some wins on the board. Around that, we need to be clear about what the AI need to do that, so it needs good context, it needs good data, uh, it needs good cross-functional knowledge, and it needs good human [00:30:00] judgment around those projects. And then you had that really good framework at the end of now, next, and future. So what are you gonna do in 30 days? What are you gonna do in 60 days? And then what are you gonna do in 180 days, which is the quick wins, the pilots, and then scalable transformation. And I think that's a really good, um, framework to really go and do the transformation.
[00:30:20] Jonathan: So we're saying to people, "Stop rolling out tools. Start redesigning work," is the, is the, is the key thing here.
[00:30:26] Danny Denhard: And obviously the tools is integrated into this. So you can, you can roll that out right at the start as well. Start and the mapping exercises, integrate the tools into that, and then those small wins will start compounding over time
[00:30:38] Jonathan: Exactly right. aim for me is, is not about just having AI in the business, it's to... The aim is to have a better business with AI helping the right people do the right work, faster, smarter, more consistently, whatever advantage we can take. That's the key thing for me
[00:30:56] Danny Denhard: Couldn't agree more
[00:30:58] Jonathan: What's the TLDR, Dan?[00:31:00]
[00:31:02] Danny Denhard: Do an AI transformation project, follow our six, six steps, iterate, share knowledge, and you'll win. Or download or download...
[00:31:11] Jonathan: think
[00:31:12] Danny Denhard: Or contact us and download it, yeah, is the other way of doing it
[00:31:15] Jonathan: that's right. I think, um, yeah, w- as we said, AI transformation's gotta start with the business, not with the tools. the best, the best opportunities are probably the cross-functional ones. So it might start departmentally, the cross-functional ones.
[00:31:27] Jonathan: And, we need to have the right context, prioritisation, human judgment. So enthusiasm is a great place to start, but you can't just go with enthusiasm. There's gotta be stuff around that, I think, make that fly. I think we've covered it
[00:31:44] Danny Denhard: I think we have. And as always, we will be sharing a lot of this in the supporting newsletter. And we will add some of the, the links, et cetera, into the show notes. So if it's difficult to keep on top of or you want a list of it, or you wanna [00:32:00] whack it into an LLM of your choice or NotebookLM, that might be a good, good place for you to, uh, to do it
[00:32:07] Jonathan: How about that? seeing you, my friend, and I will see you next time
[00:32:11] Danny Denhard: Thanks for listening today everyone. If you liked what you heard, please give us a rating and review in any of the podcast players. We have a supporting newsletter, which is aimoment.co uk that comes out on Mondays and Fridays with every pod, and we dive a little bit deeper, so please subscribe there.
[00:32:29] Danny Denhard: Have a great week, and we'll see you soon.
