Introduction

Hi, so I'm Anthony. I'm the founder of The Distance. We're a digital product company based over in York.

We've been building apps and tech for the last 17 years. I promised I wouldn't try and be too salesy, but that's what we do.

So today, whilst I've been embracing AI for the last, I don't know, 18 months quite heavily, we've been looking at how we adopt it as an agency, both a scared that it's coming for our jobs, but at the same time, how we we can really utilize this technology.

Now, my background's tech, I bleed ones and zeros, but my role in the company these days is more the front of house. I do the sales, marketing -y stuff, out here doing these sort of talks most of the day.

So I'm not an actual salesperson. I've never been on a sales program or anything else, but I've been closing for the last 17 years for our business and I've embraced AI to just make my life easier because I'm very short time.

and so what I'm just going to sort of walk you through is how we use various AI tools to just basically I've managed to work out I reckon I can close as I've closed the time to do a sales pitch or sales presentation deck from about a day or two depending how complicated the project is down to maybe a one or two hours if it's not too complex just because of the various steps we go through.

So I'll give you a quick brief run through. This is version one and you'll see version two in a second, which I'm really open to any more suggestions.

Where Leads Come From (Inbound & HubSpot)

So we're lucky our marketing does very well. We're very lucky.

But most of our leads come in from our inbound. We rank the first page of Google for the main terms. We worked hard at it over the years.

I've already signed up for Tom's product. It helps me understand where we are. Definitely get involved.

Oh, that's slightly wrong. That shouldn't be white. Those leads come

into our HubSpot and our HubSpot then we then invite them to meeting invites and we initially get get them on a call and therefore we'll send them a templated email saying here you go here please welcome here's some blog links whatever it might be fairly triple sales process right

we finally get them on a call and that'll be with me normally depending on on what we think of the

client and how we rate them or how we qualify them and we use otter to then record the calls is everybody must sit in a meeting where there's probably more note takers than them right how many people has been in a meeting recently that hasn't had a bloody note taker in right um so we use otter

Discovery Calls: Recording, Note-Taking, and What to Capture

i don't know why we actually a gemini we're a google house so we should use gemini but i prefer otter um so what we do is in that call i have a very rough idea of all the information that

i need to capture from that lead whether it be what's going to finance the project what their core priorities are. Is there a timeline to the project?

I've been doing this 17 years, so it comes quite organically to me. If I employ someone to do this instead, they'll have a proper script to do this.

But I know the things that I need to capture from that call in order for me to then make the best use of AI going forward. So whilst I'm on that call, I'll also be making some notes, and I'll just be making the key notes that I need to highlight when it comes to moving this process forward.

Turning Call Notes into a First-Draft Proposal with ChatGPT

So I then jump into chat gpt and uh i will show you i'll try i was going to do live demo but i didn't think it's worth it because it'll take two minutes to process this so i have a project set up in chat gpt i have to move this to the window um so you're familiar with chat gpt and if you're not you have projects

and within projects you can give them instructions various things so i have this project called sales as you can see here but I've set up a temporary one for today and as part of this I've set some instructions and basically I say we're gonna do some salesy stuff I've just given it some really really rough just rough outline

of what we do just so it has some context but that isn't the main stuff we're going to use for prompting but it just gives it a bit of a shape what we What we then do is I'll start a new conversation.

And if we go back up to the top.

So what I'll do is I'll upload the transcript that I'll export from Otter .ai. I probably can use some automation to do this, but I don't at this point.

I'll also then add a load of random notes that I'll have made. Absolutely me talking, just jotting down anything that comes to mind whilst I'm on the call.

So I'm getting some hints here of their intent and the relationship they're gonna be like with us. He's doing some cool stuff over here. This is about an app for scuba diving.

If anyone's into scuba diving, check out scuba, or now it's going to be called Descent. He wants to use some particular technical things in their product about matte box and everything else. So that was just basically the conversation.

So we've had a 45 minutes call, recorded it, made some notes. And then what I'll do, so it processes all that. It says, oh yeah, that all looks great.

So it thinks about it, tries to do some stuff. And what I've actually asked it to do is use our project template.

So if I go back up to this project again, you'll see that in the sources, I've got a couple of things. Now, I actually have a few more in our sales project, but I added these today. So I have basically that's our portfolio, the success and testimonials. It's just a load of case studies about the work we've done in the past for it to reference.

And then I have this proposal template. So if we jump back into that and just see what I asked it to do.

there you go so I've said here you go I've got a call recording here's some notes use our project now I have a more structured prompt to this normally but I was just doing this on the train

on the way in create first draft of the proposal so it went away and it spent about 17 seconds yep 17 seconds it's pretty quick today and to create the first draft proposal now I'll show

Before I show you what it gave me, I will show you what I gave it.

The Proposal/Pitch Template and Prompting Approach

So this is a Google Doc. I export this as Markdown normally, because AI has a preference for using files like that.

And essentially, it's a structure of a proposal or a pitch deck that I wanted to produce off the back of it.

So if I go... And you'll see that there's certain things in square brackets.

If you read in those instructions that I gave it, I told it to make note of everything in a square bracket as an instruction, so that I have this proposal document, and I want it to give me all this stuff.

So if we go to the more obvious stuff at the top,

I want it to give me a summary of the project, I want it to work out all the user personas that are gonna use this scuba diving app,

whether it's someone actually diving, wanna capture something whilst they're out at sea, or they'll upload something afterwards, afterwards or if it's someone looking for a dive center or there's various other aspects to this product. So it should try and break down all those people for me.

We then, I want it to try and work out what it thinks an MVP would be. MVP, minimum viable product, so what their first go -to -market should be and summarize that for me.

I then ask it to think about the actual technical side. So do they need an admin console? Do they need backend services?

Then speculate on what next phases are what assumptions are and potential risks one of the things that i find the most

useful though is oh i haven't got this sorry it's missing it's a business section this is an old

version um i've now got quite an extensive section here about doing some research for me and that's the bit that we save a lot of time so i have it go and do um some deep i explicitly say deep research

to keep tick tick the chat BD prompt and it'll go and analyze this product its opportunity and then try and find whatever other comparable products are in there in market they'll then be able to highlight what the pros and cons of those products are so I can name drop them and look like I know exactly what I'm talking about when I'm pitching which I black pretty much all the time

as you can see and then one of the final things it goes through it tries to

estimate it I've given it some structure to estimate stuff and then I tell it to So look at the other portfolio document I've got.

We've been building apps for 17 years. I've got 170 of them in our portfolio.

If one of our other members of staff who hasn't been there for 17 years, they don't know what relevant stuff we've got in there. So it'll helpfully quickly pick through there.

And so let's see how it's got on.

What the AI Produces: Summary, Personas, MVP, and Research

So here we go. So a quick summary, it's gonna build this scuba diving app. It highlights what the opportunity is.

This really helps when I'm selling because it replays to the customer that I've been listening and I know what they're actually trying to achieve rather than just pitching what I think they should have.

It's then broken down the personas.

And without one of my competitors, I don't think he's in the room. He actually has an office just there. I'm giving him all my secrets.

So we talk about success. And so when we start a project, we always try and use those metrics for demonstrating what we're going to achieve over the course of the project.

So we always when we talk about the individual personas, we talk about exactly what success looks like for them So it does a great job of that nice for clean bullet points and

Then it gets down to the MVP It's broken that down highlighted the conversation looked at what we said the priorities were You'll see that I then gave it

The client came back to me and then sent me all this long email saying oh here are the priorities priorities.

So what I did was I just gave it all that and said, revise this. So again, it did. It revised the summary, revised the MVP down.

If I don't think this research was good enough, so for example, I did one yesterday and the client had sent me three specific examples. The research chat GPD did didn't include any of them.

So I was able to give them back, added them to the mix, helped me iterate it. And then we get down to sort of the typical stuff.

Estimates, Pricing, and Ongoing Costs

if we go back up to that pricing so it's even done a pretty good job at estimating it so um here we go so it thinks it's going to cost 75 to 100 grand on 115 to 200 grand that's a pre -2026

price really we now use ai and everything we this project ended up being i think for their mvp build we were going to do it for i don't know our sweet spot's about 60 to 90 grand for an mvp for a really good product and so but that is easily what we probably would have

charged them in 2019 say and then it even starts to break it down let's make this easy let's structure it and it knows all this because it's also got the context of I don't know a year's worth of us having chats with chat GPT so it knows how we work the way we do things these prices it's estimated some of it

it will have used some estimating we've given it in the past but most of it is actually it's just calculating itself I've noticed there's a big difference in the last 12 months how better it is at doing that it must just have more data so for example it these kind of costs it'll tell me based upon the hourly rate

probably as well yeah there you go it's even broken down what their ongoing costs are so it thinks what is their multi their maintenance costs are super such hosting costs so that's version one that's where we are now so whilst we're

building lots of cool stuff with ai my actual use of ai is still quite naive so most of this doesn't seem probably that amazing to most of you probably all quite versed in all this so yeah so i'm still

What’s Next: Improving Automation and Tooling (Claude, MCPs, Gamma)

working on what we look at do next um i like to get hubspot we just upgraded our hubspot account so that'll do a bit more enrichment of those accounts so we can do a bit more validation of

who they are up front we could probably use um gemini in our google to actually create slightly slightly favourable emails, so it'll tailor them a little bit more based upon their ask, maybe pull out a case study that's relevant to the thing that they've asked for.

We'll still use Otter. We're going to move everything to Claude.

Everyone's moving to Claude, right? That's the current trend. It's just infinitely better.

For us, using MCPs, which I won't bore you with today, that's a different topic, we can just plug into everything. We have an MCP for our board, and it plugs into forecast and zero and everything, and it just just gives us access to everything all conversationally.

But I reckon I can use Claude, a bit more automation to generate this. Maybe use something like Gamma.

Anyone familiar with Gamma? It's like one of the hottest AI presentation things at the moment. There's loads of them now, but probably do it in pitch decks of some sort.

And then what we need to do is continue in the follow -up email workflow. We have some of that in HubSpot, but it's not very AI yet. It's quite manual steps.

And yeah, hopefully it'll save me another few more hours. I don't know how.

Conclusion

And yeah, that's it. that's what we do thank you very much