EPISODE 2: Why the AI Guy Wants Your Timesheets

Full transcript

Auto-generated transcript, lightly formatted for readability (punctuation and paragraph breaks added; wording is verbatim).

Callum: Hello, welcome back to the STANDOUT Agency Podcast. My name is Callum Healey.

Gareth: And I’m Gareth Healey, and we are Agents of Change, which is the URL of our website, coincidentally, and that’s turned into quite a nice podcast opener without us realising it. Who knew? So that’s great. We have a topic that isn’t usually an exciting one, which is timesheets in agencies, which, Gareth, you’ve had your fair share of experience with.

Gareth: Yeah, totally not an exciting one. Always provokes a lot of interest and debate. So do you want me to lead off a bit?

Callum: Of course, you’re the expert. Let’s do it.

Gareth: So there’s a little bit of a story here, actually. I did my degree at university, I did a placement in the third year of my degree, many years ago, obviously, and part of my placement was to manage the timesheet recording system in a software house. And at that time I thought, well, that’s not particularly an interesting job, but I’ll do it for the degree, and I’m never going to touch this timesheet thing again. Here we are 30 years later, and I don’t think a day has gone by without talking about timesheets, because of course I got into the agency business, and timesheets have been a factor – love them or loathe them, timesheets are around in the space. And what I found interesting was 12 months ago, when we started working together, you, Callum, were completely oblivious to the concept of timesheets.

Callum: And shocked, actually. Timesheets, really? Why? What, people recording time?

Callum: Yeah, I was thinking, is it not value-based pricing? You know, if something gets to the client quicker, is that not more valuable to them? But of course, hourly pricing is something that is the most easy to control.

Gareth: Yeah, and there was another shock more recently, because in your interactions with agencies where you’re helping them adopt and use AI, you’ve had a bit of a revelation in the interactions with some of the agencies you’ve had so far. There’s been a theme.

Callum: Yeah, basically every single agency that I’ve worked with has been trying to figure out a way to delegate that task to AI. And why wouldn’t you want to do this? I think a core philosophy with AI: the repetitive, laborious, manual work, you want it to do the boring tasks, whereas the people in your team can focus on that human-to-human interaction. There are classic clichés in terms of what human intelligence excels at, in terms of strategy, judgment, creativity – things like this. I think in general it’s specific to the individual. If you’re someone like me who loves analysing things, I’m always picking apart processes, trying to make them more efficient – not just in business but in my own life, my health, etc – then you might want to do more of the analyst work even though AI is very good at it. I found that early on in the business that we’re a part of, constructing things like websites and PowerPoints all the time – not my strongest suit, nor do I much enjoy it – and so I want to delegate that to AI, most of that, and set up those systems. Someone could have the complete opposite, where they love graphic design, they don’t like analysing things, and it’s like, let’s just get AI to do the things that we don’t want.

Gareth: Makes sense. And hopefully that won’t come back to bite us, in the sense of AI becoming resentful in the future.

Callum: No, fingers crossed.

Gareth: Fingers crossed, yeah, that is the problem. But we’ll see. For now we’ll let it do the work we don’t want to do. So the title of this podcast then is “Why does the AI guy want your timesheets?”, and that has come as a bit of a shock to us. But I think it’s particularly a shock because the fact that timesheets are a prevalent feature of requests regarding AI – you just explained some of the reasons why, because it’s laborious work, but also there’s lots of people out there, agency owners in particular, that frankly have never liked timesheets. Nobody likes filling these things in, even agency owners don’t like filling them in, and even when they’re filled in – and we’ve got some data – it’s people that rarely or often don’t use it. So there’s a strain of thought out there, definitely, and it will be growing, to say, you know what, guys, who cares about these timesheets anymore, because we’re going to be using AI, and tracking individual human time will be a thing of the past. If we ever needed to do it, we don’t need to do it now, going forward.

Callum: Well, I think in relation to that, I have a client that I work with who do media planning, and they’re on a commission-based pricing model, and therefore have never done timesheets. In our work together it’s something that I’ve suggested they actually incorporate into their work, which you can imagine their faces in the room when I suggest introducing timesheets. But ultimately, I mean, you laid it out perfectly, and our friend and CFO Alfie – if he’s listening, shout out to Alfie – he was saying how important timesheets are in the case of how agencies are: it’s a knowledge-based industry, and people are the kind of raw materials, and time is the cost. In a way, if you’re able to work out what your cost is, you’re able to calculate your revenue, and then you can calculate profit. But if you don’t know what the cost is, then that’s almost half of the equation that’s completely up in the air.

Gareth: So I think it’s always important to incorporate that.

Callum: And it’s actually over half the equation, because for most agencies these days it’s 60 to 70 percent of their costs are in people, are people-related costs. So it’s a huge portion of the cost of the business, to have no data, to have no real insight into how those costs are being deployed. And I know that sounds quite – these are human beings at the end of the day – but that is the reality of it, that’s a cost, we’re deploying 60 to 70 percent of our costs in the agency deployed in people, but to not fully understand what they’re doing and the productivity from that cost has always, in my mind, been a very strange decision for many agencies, and also a huge gap in the business.

Gareth: Completely. And I think that if you were to run a business that was based on products rather than services, like a car manufacturer – was Alfie’s analogy – the car manufacturer would know the price of every nut and bolt, no matter how small and insignificant it might seem, everything is meticulously priced. So any upgrades to the car model, or repairs, buying the car, selling the car, they’re able to work out exactly how much money they can gain off those kinds of transactions. But with AI it adds a separate layer, and it’s the fact that speed ultimately is AI’s main value – of course it allows you to amplify your current expertise and the expertise of your team members, to make the value of your work stronger as well, of course, but ultimately you get it done quicker, you can do more. And so this has implications with time tracking, because when a process is very quick, it’s harder to control. And so having an automated system that has all the formulas in place, has perhaps your initial time that it took for a workflow versus AI-enhanced, you can actually see the amount of hours that you’ve recovered. And it’s almost the most important build, I’d argue, because when you start to incorporate AI there’s a lot of cost there in terms of training and subscription costs, and so we need to be able to prove – agencies need to be able to prove – how AI is affecting their capacity, so they can then change their pricing, architect their model, their capacity more effectively, so they can actually reclaim some of that time that is saved. It’s human nature to finish a task in half the time – perhaps you then go, okay, I’ll just go on LinkedIn for a bit, or I’ll chat to my mate at the coffee machine, I’ve done my work for today – it’s a case of intentionally reallocating that time somewhere effectively, so you can reclaim the margin rather than just doing it in a shorter period of time.

Callum: So if we’re trying to deploy AI within an agency, and specifically looking to deploy it to help us manage our costs better, to help us manage our capacity better, in terms of what we’re doing to manage and assess and analyse our productivity, data is needed, isn’t it? And if a large part of that process at the moment is human labour, we need the data of the human labour to feed into it, don’t we?

Gareth: Very well said. I think that is the core philosophy to understand about AI: data is the value. I did a psychology degree, of course, many may know, it’s on the website and whatnot, and we kind of modelled the AI brain after the human brain – it’s got many similarities. The brain is the only form of intelligence we were ever really aware of, in humans and animals, so it’s like, okay, let’s take some of the inspiration from this kind of biological hardware, if you will, a brain. On its own, when someone is born, or perhaps if someone’s memory was completely wiped, it’s a clean slate, and the environment is how the brain learns. And so for your team members, they learn what your agency does and how to do certain workflows, and their current values and the brand voice, things like this. AI doesn’t have any knowledge of that, because it hasn’t had that environmental training. But instead of the environment, for AI I call it data – data is that value. So what makes it special, what makes your agency special, everything that you could possibly teach an employee about a task, AI should know that, because you want AI to work for you rather than just be a blank-slate, off-the-shelf product. And so you need to track the data – unfortunately timesheets is a way of doing that. Actually filling them in is the thing that AI can’t do, unfortunately – that’s the one thing where it’s a behavioural factor, where it’s just up to your team members to fill in the timesheets.

Callum: But I think this is downstream from education – once people know why it’s so important.

Gareth: Yeah, that’s massively important, yeah, because there are – you’re right, a lot of people need to reflect on and complete their own time.

Callum: There are AI-fuelled and AI plugins to most of these, numerous systems now, of course, aren’t they, they’re trying to help – using AI to help people fill them in, by tracking their activity online and what they’re doing, and linking it to diaries, etc, so there’s more help than ever before.

Gareth: When I – again, I’m showing my age – timesheets at the very start of my agency career were completed with a pen and paper, and somebody data-inputted them. There was a worse job than completing timesheets: it was completing other people’s timesheets and inputting them into a computer. So, wow, we’ve come a long way, but AI is helping to do that. And when you’re talking to, as you have been, agency owners – what’s the focus? Clearly timesheets, we’re doing this episode because timesheets has been a topic – what are they asking for, then, without divulging obviously any client confidentiality? What are the requests that you’re getting, and why are they interested in this, with using AI?

Callum: Ultimately it’s visibility, I’d say. It’s to have complete control and information about how your agency is currently performing. All of the data is usually tracked by these tools – I’ve learnt so many in the past six months, like Synergist, CMap, Teamwork – there’s so many you can rip off, and usually all of them have an API, which is just like online plumbing, you know, you can get an API key, AI or an automation can then access that data, as long as you authorise with your username and password and so on. Some of them don’t allow the API if you don’t subscribe to the most expensive subscription, which we’ve almost had a theory that – does it even exist, this subscription, this enterprise plan, or is it like a price-anchoring situation? But what I’ve learned is the data itself is not that hard to get, because, like we said, it’s behavioural – people fill in the timesheets, they answer a few questions. I’ve never filled in a timesheet personally, but I’ve seen timesheets of my clients, of course, but the data is all – it’s nice and organised – but some of these AI tools, where it’s like, you know, we’re – let’s say Teamwork – we have an AI for our timesheet system – they’re rubbish, because ultimately all of the most popular AI models, Claude, Gemini, ChatGPT, the whole company is putting as much money and time and energy into just AI, whereas some of these timesheet tools, they’ve got their actual system to manage rather than just creating an AI system on top of that. I’ve been working with a team recently where they had two streams of data tracked, where it was resourcing hours – you know, the amount of hours they put to a certain client or workflow – and how much the client was paying them for that workflow or time period, but those two streams of data weren’t allowed to be compared on the app, on the system. And so, unless you want to get your calculator out with a piece of paper and manually do all of these calculations, AI can help you with that analysis layer. It’s like, it’s got the data lake, which might not be very organised – it’s an automation, ultimately, and it can come in many formats, but it’s dependent on the type of things that you want to track. So some software might track rate cards of your clients, like I said, resourcing hours and how much the client paid you, or the cost, which is hours in this case, staffing pinch points, you know, holiday hours, each person’s name in your agency, what workflows they do, how often they do them – this can all be tracked, but then it’s about AI picking apart what it all means. So to answer your question, there’s three layers to most AI systems, but particularly in timesheets in this way, and it’s extraction of the data – so that’s with an API, or even some automations, the ones I primarily build can be started as soon as a spreadsheet is exported from the site if you don’t have access to that API. So it’s extraction – AI gets the data – then the automation is programmed with deterministic formulas and code, which is something I’ve learned, AI by its nature, as many people know, hallucinates, it can never be truly reliable, and so you can’t let AI itself analyse the financial truth of your agency, because that can have horrible implications. And so it would run through code for the numbers, through formulas that we set, or the teams I work with set, because these are the formulas they want. AI is then the reporter of that information. So it’s extraction, analysis, and what do you actually – what insights you want to get – and with AI it’s almost like the limits are endless on this, you can describe the system, and as long as it fits in the layers of extraction, analysis and reporting, the system can be made in that way. You get the numbers, you get the analysis, and then it’s delivered at a certain cadence – so it could be a custom AI system that you query freehand and go, you know, ask a question of your choice, and it will give you the answer, because it will go and access the data, or it could be reporting on a cadence, you know, email, Slack messages, team reports, to leadership and whatnot.

Gareth: Makes sense. What surprised me, actually, because you’ve been dealing with these clients and working with these clients – was the fact that when you came back and talked about this, it didn’t surprise me it was an issue, because we’ve said timesheets, time recording, has long been an issue and a topic of debate in agencies. But what surprised me was that these various and numerous software systems, many of them very expensive, don’t – they have the reports, but the reports either aren’t flexible enough or aren’t giving people the data that they want. Surely some of the particular bigger ones are able to furnish the agency owner who pays for it and puts the data into it with the right report, flexible enough to compare things – but that doesn’t seem to be the case.

Callum: That’s the constraint, yeah. And so you can develop a system of your own, a proprietary system, rather than begging the provider to give you this niche form of analysis that’s only relevant to your agency – you can create a system that you own, you can update over time, it’s all private, all of the logic is completely your design, and then you don’t have to rely on these rising costs, because you built it yourself.

Gareth: And potentially that makes sense, and I get so surprised, but I’m thinking now, I have a number of conversations with agency founders in the past that have used some of these systems – without naming names – and saying, we really want our report to do this, so we want it to do that, and it doesn’t quite do that, which is part of course of a lot of this work. But potentially, with that extraction and analysing the data, potentially the agency could move their data input, their time-recording software, to a different provider if they wanted to, for cost reasons or for any other reason.

Callum: Yeah, and that goes back to what we said about the data being the value, because as long as that’s thought about carefully and constructed and developed over time, perhaps it’s the source of truth for everything to do with the context of your agency. You don’t have to rely on these different providers, because, in that analogy, the providers are – they’re giving hardware which processes the software, and it’s like having a three-year-old’s brain as the data, but compare it to Einstein’s brain – I’d rather plug Einstein’s brain into a system than a three-year-old’s brain for my company, yeah, it’s that kind of thing.

Gareth: And in the form of timesheets, what does your perfect timesheet system look like? It’s about defining that first.

Callum: And I think extraction, analysis, reporting – it’s about filling in, okay, what’s the perfect scenario in each of those three stages. I guess it’s exciting because people can’t wait to not have to deal with them.

Gareth: Yeah, which leads me on very nicely to a key point really that we must talk about, which is, you know, people will say – listeners, I’m sure, will be thinking – well, you know, whilst I set this up at the start, inevitably there will be a point in the future where we would probably want, or probably embrace the fact that we’re not recording time, because AI is doing a lot of the work.

Callum: Yeah, and that’s – I don’t think either of us have got a crystal ball, but I don’t think either of us are saying that that potentially is not going to happen – that we might not – AI might be doing such an amount of the production work that 60 to 70 percent of an agency’s, a future agency’s, costs aren’t in people, they might be significantly lower. It might be 30 to 40 percent of the agency’s costs are in people. Tokens might become the new currency.

Gareth: Great, absolutely. So if that’s happening, again, for people – I don’t think that might well happen, but I think my take would be – interesting to hear what you think, Callum – when I’ve been talking to my agency clients on a broader advisory role, is that there is a window of opportunity here, and it is an opportunity to understand in more detail than you’ve ever had before exactly what your productivity and capacity and people-based hours are, what inputs are in your agency, where is the time and therefore where is the cost currently going in your business, and is it going to all the right, effective places, or is some of it being wasted, frankly, or is some of it being spent on things that don’t need doing, or some of the time has been spent on clients that aren’t contracted for that amount of time, and therefore we’re over-servicing. So there’s a window of opportunity that’s always been there, but I still think it’s very valuable, rather than ignore that window, to understand where things are now, because – interesting your view – because when in the future AI is used even more in business, we will have a base of information to understand: this is the type of work we did with purely human labour, and this is the amount of hours, this is what the time cost us, this was the inputs that went into it. Now we’re doing it a different way, and this will be a transition – it won’t be overnight, whatever happens, it won’t be suddenly we’ll wake up on Monday morning and start working like this, it’ll be a transition, it could be a very short transition, but in the future we might use AI to do a lot of those inputs, but we have a knowledge base, a data set, which says this is how our costs and the inputs to our business used to look, now with AI this is what it looks like now, and we can compare the two. And what I’m saying is, you can then – you’ve then got a very much a clear, data-backed view on how you might need to adjust your price.

Callum: Yeah, and I think the almost dystopian future of that, which isn’t necessarily a bad thing, would be AI constantly recording people’s computers to see how long the tasks took, and then just automatically calculating the time of that. I do think no matter how far technology advances, time is never going to change – it’s still very valuable to know where you’re using your time day to day. It’s like, you’re awake for 16 hours a day, it’s like you have 16 tokens, where are you putting those chips? If you’re on your phone all the time, and you look at your screen time and go, I spent five hours on my phone each day, that’s good for you to know, because you’re like, maybe I shouldn’t do that, do something a bit more productive. The same in business – if you know, it’s just good to know where you use your time, because ultimately it controls everything, doesn’t it?

Gareth: No, totally, I love that, I like the “hours as tokens” concept, and, you know, this again to people in teams, we’ve got to ensure that we’re not ever being seen as being Big Brother in this and checking up on people, because really that, for me, has never, regardless of AI, been what this time-recording stuff has been about – it’s not about checking up, it’s about using time effectively. And we’ve only got 16 hours a day awake generally, but people are with us, that we employ in our teams and agencies, for seven, seven and a half hours a day – there’s seven, seven and a half hours of tokens that we’re actually spending. As an individual, it’s a human being spending them, all the time, every waking second you’re spending this token of time, and we want to know that token’s being spent effectively – we’re not trying to make everybody into a machine, but that is a mindset that – if we’re spending that, then we need to get the most out of it, don’t we, don’t be wasting that time, that token, or however you want to call it. And we’ve not wanted that in the future, certainly in the past, certainly don’t want that in the future, when a client comes to an agency owner or principal and says, hey, the work that you’ve been doing for us, whatever that work might be, for the last couple of years, you’ve been charging X, and you’ve been charging X based on you’ve told us that that takes a number of people and most of their day, whatever it is, and now you’re using AI to produce a lot of this work.

Callum: We know that as a client, or I know that if we had the answer to that, I don’t think we’d be sat here doing this podcast – I think today I think we would be maybe setting up our own agency, or certainly spending more time – we don’t have a direct answer to that.

Gareth: It would be ironic if your life’s task was to make a timesheet system.

Callum: Yeah, no, let’s not even – if it was possible, we probably wouldn’t do it.

Gareth: No, I don’t believe anybody has, as an agency owner, has a very clear, 100 percent answer to that client question. But I guess my point is, if you have data – if you’re guessing, if you’re just using it, trying to answer that question without any factual evidence and without any data, not necessarily directly with the client, but before you have that question with the client, when you’re thinking about how your agency operates – you need this data, don’t you?

Callum: Yeah, and I think if you had a client that had to let off an agency, and they had two agencies working for them, and the discussion was about, okay, we’ve seen AI is speeding up some of your processes, you’ve got it on your website, you say you’re AI-enhanced, we want to know what’s going on here, like how much time is it really saving you, because we’re paying you X amount – like you say, we want to know how we can renegotiate this contract, we’re struggling right now, we have to let an agency off – so if you can give a clear example, great. Agency A could go, we don’t have a clue really, we know it helps, but we don’t track it, we just kind of use it, and somebody’s using it all day practically over there, this person’s using it again – it could be completely systemised, like they have a great setup with AI, agency A, but they’re not tracking objectively how much money they’re making or saving by incorporating it compared to what they used to. And so agency B, they might not be that AI-enhanced, but if they know exactly how it affects their margin, they say, now we’ve embedded AI, we are saving around 30 percent of the time that we used to take, we’ve made a 30 percent reduction in time spent, and overall for you in a rate card etc, whether they choose to give that knowledge to the client, that’s another thing, but at least they know that. And perhaps it’d be a case of, you know, we’ll disclose that we actually save 10 percent, and we’re like, yeah, we’ll give you 10 percent – but my original analogy was going to be, we have a 30 percent reduction, we’re willing to give you 15 percent of that and reduce the cost and reduce the price by 15 percent, they’d go, okay, or we can do more work, we can take that 15 percent and do some of the other work that they were doing – true, yeah – so it’s not even about reducing the price, it could be keep the price, but we’ll do 15 percent more, or we’ll do 30 percent more, but it won’t cost you 30 percent more, because you’re saving money between just using agency B and not agency A.

Gareth: The great salesmen out there, they know their craft, but without the data they don’t know how to use those words for it to sound, or influence an agency or a client, to build a commercial proposition.

Callum: Exactly, and so that’s why it’s so important to track, so you have the truth of what’s going on.

Gareth: Great, so I think that was a really interesting discussion, but if we wanted to summarise – not summarise the conversation, but – what would we want agency owners listening to this to go into their businesses tomorrow morning, or whenever it is, and do?

Callum: And for me it’s – if you are not treating time recording seriously, and you never have, please don’t see this as an opportunity to feel vindicated that AI is here so we don’t need to do that anymore. I’ve got the opposite direction – I’d use it as that window of opportunity to record and start to get your data, in terms of human input and human hours, human labour – get that data as a data set right now, so that you can use it in the future.

Gareth: Anything to add to that?

Callum: Definitely – I mean, I’d say that timesheets have never been more important to fill in, because of everything we’ve discussed, you need to be aware of exactly the effects that AI has. It’s also never been easier – so that dreaded meeting you might have scheduled about discussing timesheets and getting people, you know, incentivising, please fill this in, which is still very important, and AI can’t do that initial input part – but if your team knows and understands the importance of filling it in, but perhaps they get some of that profit, they get some of the profit-sharing, or even simply a free lunch or some sort of recognition – there’s a reason we get this much margin, we’ll give you this benefit, there’s things like that. But ultimately AI can take that off your plate, and so it’s a bit of a no-brainer for me if I was running an agency.

Gareth: Good call, thanks Callum, enjoyed that, as always, look forward to the next one.

Callum: Yeah, see you in episode 3. Thanks for listening, guys, see you later.