← Video Library · Gamma
Transform Your Presentations with AI (Claude & Gamma)
Published November 22, 2024 · 99 views on YouTube
Key takeaways
- The A4 framework orders AI adoption as Accelerate, Amplify, Augment, then Automate, saving automation for last rather than first.
- Writing Claude's presentation output in markdown, inside a code block, preserves headings, bullets, and structure when pasted into Gamma or Google Docs.
- In Gamma, choose 'Preserve' rather than 'Condense' or 'Generate' so it keeps your Claude-written content exactly as written instead of rewriting it.
- A Claude project can hold a company's background, goal, and a numbered instruction list so a multi-step prompt sequence can be reused over and over.
- Claude can turn raw financials into an interactive React graph that can be dropped straight into a presentation.
What is the A4 AI integration framework?
Bruce Fleck’s A4 framework orders AI adoption into four stages: Accelerate (do something you already know how to do, faster), Amplify (improve the quality or reach of that same work), Augment (bring in frameworks and skills you don’t have by pulling in proven methods and letting AI apply them), and Automate (only after the workflow is refined and proven). Automate is deliberately placed last, not first, because Fleck argues most people try to automate a process before they’ve proven it manually with AI assistance.
How do you use Claude and Gamma together to build a presentation fast?
In the demo, Fleck used a Claude project pre-loaded with the A4 framework and a company scenario (a fictitious HVAC software company, Comfort Tech, needing a growth-strategy presentation for leadership). He told Claude to proceed through the four stages step by step:
- Research and data gathering — Claude generated prompts to analyze the market and competitors instead of doing the research by hand.
- Strategic framework (Augment) — Claude suggested established frameworks (McKinsey’s Three Horizons of Growth, Kotter’s 8-Step, Balanced Scorecard, the SCARF model, and the Four Actions Framework) and combined a belief-shifting model with a strategy model to address a leadership team not yet on board with automation.
- Presentation draft (Amplify) — Claude wrote a persuasive narrative structure and produced the full presentation in markdown inside a code block.
- Build in Gamma — He created a new Gamma presentation, pasted in the markdown, set it to preserve the text, chose AI images of business professionals, specified 20 cards, picked a custom theme, and let Gamma generate the deck.
The full build, from research prompts to a near-finished 20-card presentation, took just under 20 minutes in the session.
Why write the presentation in markdown before pasting into Gamma?
Markdown makes slide boundaries and formatting explicit and unambiguous to Gamma: a single hashtag becomes the title, double hashtags become a heading two, and bullet points and numbering carry over correctly. Fleck noted the same trick works in Google Docs — enabling “paste from markdown” under Tools and preferences in Google Docs settings applies headings properly on paste instead of dumping in unformatted text.
Which Gamma setting keeps AI from rewriting your content?
When pasting text into Gamma, choose “Preserve” rather than “Generate” or “Condense.” Preserve keeps the pasted content exactly as written; the other options let Gamma rewrite or shorten it, which caused problems in earlier attempts where Gamma altered content that had already been carefully written in Claude or ChatGPT. Fleck also pointed out Gamma’s image-model options during the build, noting Flux is strong for faces and Ideogram is useful when images need to include text.
Why does the A4 framework put automation last?
Fleck compares it to learning piano: you don’t start with a Rachmaninoff piece, you learn the basics first. He argues clients feel less pressure and build confidence faster when they learn AI’s strengths and limitations, then basic prompting, then advanced multi-step prompting with frameworks, and only automate once a workflow is proven and repeated. He noted that a working multi-step prompt saved inside a Claude project (similar to a custom GPT, but private) is itself “half of your automation,” since it can be reused repeatedly, for example for email nurture sequences, landing pages, content pillars, or micro-course builds, without needing full automation. He said the goal of his coaching practice is to give tech-interested but not tech-savvy business professionals an AI foundation and basic skills before they take on heavier automation.
Full video transcript
Hello everybody and welcome to the Scale with AI Summit. My name is Kelly Tenny, and we are here with another amazing hack session, and today I have Bruce Fleck, who I’m proud to call a personal friend of mine and definitely someone I always think about when I think about AI, especially when it comes to hacks, because Bruce always has something really cool up his sleeve, something new that he’s doing. Bruce, I don’t know if I’ve shared this with you, but you really are someone that I know is always forward thinking, testing out new things, so I couldn’t think of a better person to have here on our hack session today. So welcome. Thanks so much, and it’s always a pleasure working with you, Kelly. You’re such a warm, thoughtful, innovative person, so it’s always fun to connect with you. So I appreciate it, you’re on the cutting edge as well and out there spreading the good word, evangelizing for all of us about what we can do with AI to improve our lives and improve our education system, so appreciate it.
Yeah, thank you, and before we dive in, I know you’ve got some really cool flows and methods to share with us today. I want to share a little bit about Bruce for any of you who may be catching this hack session and don’t know about Bruce or don’t know a lot about him. Bruce has coached dozens of professionals, developed proven methods to implement AI for maximum impact, and his deep knowledge across technology, psychology, business, and coaching really gives you that unique insight into the challenges of integrating AI into daily work. Every time we talk, your wide range of knowledge across different industries is always something I’m impressed with, and I think that’s one of the reasons you’ve been able to not just harness AI but really have a deeper understanding of the way things can work because of the depth of knowledge you have across certain industries.
So thanks. Yeah, I have a pretty varied background: engineering school with electrical engineering and computer science, then an MBA in marketing, worked in several different industries — telecom, CAD software, internet development — then in medical devices for a number of years at big, small, and medium companies, then pursued a PhD in psychology, integrating learning, technology, and psychology, all before this AI revolution hit us. So it just prepared me perfectly for what’s happening right now. Just a little bit of dumb luck, but it’s all working out.
Sure, it’s more than a little bit of dumb luck. With that, I want to ask you — I know you’ve been working on something inside of your business, something you’re using with a lot of your clients, called the A4 AI integration. Can you share a little bit with us what that’s all about? Sure. It comes back to when we think about two things: productivity and how we manage our workflow. A long time ago in my MBA, my two favorite courses were a business strategy course and an operations management course. They seem like very opposite ends of the spectrum, but when you think about businesses, they’re made up of three primary things: products, processes, and people. In order to get the most out of people, we’ve got to have good processes — that’s what operations management was about. And when we think about thinking strategically, how can we make the most impact?
Over the years I’ve come to the realization that I don’t care anything really about productivity systems or methods — doing things efficiently, that’s great, having to-do lists, getting things done, setting priorities, that’s great — but when it really comes down to it, what is the impact that we’re making? So I measure everything in terms of impact, and that’s what led me to this A4 AI integration. As I was learning AI from you and Darby and other people, you were instrumental in me getting my feet wet and getting off and running with the course we did on Change Makers. I saw that I could begin to accelerate the work I was doing, not only do it faster but do it better and have greater impact by amplifying it. Then we learned about bringing in frameworks that allow you to augment what you’re able to do — skills you didn’t have but somebody else has already proven, you bring that in and use AI to help you. Then we all thought, oh, this is going to automate everything, and we tried to automate stuff and it was really hard, so I think automate is really the last thing you should be doing when you’re looking at your work.
If it’s okay, I could share my screen and share what this framework looks like, and then we’ll run through a demo example. I would love that, and I love your take on the idea of automation. I think there are so many AI experts out there talking about automating with AI — in fact, that’s what the majority of people talk about — so I’m really interested to see what you have to share, because I think that’s the thought disruption: wait a second, what if it’s actually the last thing you think about doing?
What I’m sharing right now is a tool called napkin.ai, and the idea is pretty simple: a lot of times when you have these brilliant ideas you’re out with somebody and you sketch stuff on a napkin, so that’s what this looks like, like a piece of notebook paper. I came up with this idea of four different areas: accelerate, amplify, augment, and automate. If we think about productivity or efficiency, that’s where accelerate and amplify come in, and then you need to increase your skill level — being able to prompt and understand systems, how to combine different systems — to begin augmenting and automating, where you can have much greater impact.
This is the definition of accelerate: you’re doing something you already have the skills to do, but you’re learning how to speed up the process. With amplify, it’s something you already know how to do, but you’re improving the quality or improving the reach of what you’re trying to do. Let me show you one cool thing in napkin — I didn’t have a graphic here for visualization, so all I have to do is click this little lightning bolt and it’ll generate some suggestions for us based on those words. I’m looking for “augment” — how does that look? Wow, that’s amazing. I can generate more if I’m not liking that, or grab that one and pick a different color version. Very cool.
And like I said, automate is really the last thing we should be doing, and I want people to think in terms of not having to automate the whole thing — you can start by automating pieces, and you shouldn’t automate anything until you’ve refined and proven the process of your workflow using AI. Once you have that, you can start automating different pieces. So that’s the framework in a nutshell.
What we’re going to do as an example is one of the typical things you do in a company, whether you’re self-employed or work for somebody else in a role like marketing or finance: you have to solve a problem your company is facing and make a presentation to leadership to get funding and buy-in. I’ve got a fictitious company here called Comfort Tech, a software company that sells back-office software for HVAC companies. My role is Director of Growth Strategy, reporting to the CEO, and we’ve got a problem — we were growing well before COVID, flat during COVID, and we’ve had trouble getting back to our pre-COVID growth trends. We’re trying to figure out what to do, and we’d like to secure funding to change our marketing and sales enablement approach. We’ve got about 45 minutes to discuss that with the executives and come up with a plan.
If I go over to Claude — and yes, I’m going to run this project right here, which will be a little bit of the automation we’ll talk about later — I’m just going to grab this information about our company and what we’re trying to accomplish and drop it in, and let’s see what Claude can do. I’ve told it what the problem is, and now it’s going to walk through the four stages of AI integration. First it tells us how long this would take manually: we’d start with research and data gathering, do some initial analysis, develop a strategic framework, then develop and refine the presentation — this would take two to three weeks of normal business time. Let’s move to the next step: what would it look like if we could accelerate these processes?
I told Claude to give us prompts on how to accelerate the process, so rather than doing all those pieces by hand, we use AI to help with research and gathering — here’s a prompt: analyze the HVAC contractor management software market, include top competitors, focus on companies in this revenue range. These are prompts that would help at each stage — research gathering, initial analysis, strategic framework, and presentation. That’s how it would help us save time. Now let’s move to the next phase; I’m just telling it to proceed. Here’s how it would amplify the quality and impact of the presentation — enhanced data insights, and a prompt to help write a compelling narrative structure, and one for persuasive argument building that looks at each of the roles. That’s how it would amplify it.
Now let’s see how we could augment this. Here are different presentation frameworks that can be used — McKinsey’s Three Horizons of Growth, Kotter’s 8-Step, a Balanced Scorecard. I need something that uses a strategic framework to analyze the problem and options, as well as do some belief shifting with the leadership team, because they’re not on board yet with using AI and automation for marketing and sales. Let’s see how that works — this talks about a SCARF model for doing belief shifting, and the Four Actions Framework. I like that — combine those two approaches and proceed. It tells me how to analyze it and specific belief-shifting elements that will help me make the presentation.
Now I told it to make this presentation in markdown in a code block, so this is what it looks like — a code block in markdown. I have this for a specific reason: when I go to put it into a presentation tool like Gamma, it can read this appropriately. I would normally spend a little time going through this to make sure it fits my story exactly the way I want, but I’m not going to do that here because of time — I’m just going to copy that and go over to Gamma, and we’re going to show you how it automates a presentation for you. Gamma is one of the sponsors of our Scale with AI Summit. I’m going to create a new one and paste in the text — all of that markdown. I want it to preserve the text, I don’t want it to change the text. I want to use AI images of business professionals, and I want it to be 20 cards because I had 20 slides. I got 20 cards, and we’re ready to go — watch the magic happen. I need to pick a theme — I have some custom themes here, I’ll use this one — and we’re ready. It’s starting to generate the presentation, this is near magic in my mind.
While it’s generating, I want to ask: I know you kept it in that markdown format — why did you keep it in that format versus just having it write out your slides and sticking it right in Gamma? In markdown it’s very clear what is a new slide, what are bullet points, what’s numbering — all that formatting comes over perfectly into Gamma if you use markdown. If you wanted to grab this markdown and paste it into a new Google Doc, there’s a setting in Google Docs — I’ll do a new Google Doc — there’s a setting for when you do a paste, “paste from markdown.” You need to go into your settings, under Tools and preferences, and enable markdown, and then when you paste from markdown it does all the headings properly — heading two here, heading one for the title — one hashtag for the title, two hashtags for heading two, and all the bullet points. It’s a great way to preserve formatting when going back and forth into Google Docs or Gamma.
I love that, because that’s my biggest issue with Gamma — you just solved one of my biggest pain points. I find it doesn’t quite capture the headings and subheadings the way I’d like, so I end up editing it a little, but I’m just pasting from a straight Word document, not markdown. I’m excited to give that a try — another hack within a hack. Yes, I watched tons of YouTube videos from experts out there, and this was a hack I got from somebody else — I’m not the genius, I’m just the little bee that flies around all the different flowers and collects little pieces of pollen and remembers them.
I’m seeing a lot of differences, and I bet markdown is the key to why I’m seeing differences in what Gamma usually generates for me. Yes, that’s one thing, and then your custom themes with your brand colors, light or dark, and different fonts. As I was doing this I was also picking the images I wanted, which were AI generated, but you get to choose which model to use — they’ve got several different image models to choose from. One of the great things about Gamma is you can change the layout if you don’t like the previous one — there’s a “regenerate” and “edit with AI” — if I wanted to make this more visual I can click this and it regenerates in a different format, or go back to the original. These are all AI-generated images. Very cool, I’ve already got something very workable.
I also wanted to show you one other thing: ChatGPT just released a new search option — if you’re on the Pro plan you’ll see a little globe icon for search. I put in this prompt: what are the top seven trends negatively affecting online marketing in 2024 for B2B SaaS providers like this company, and it gave me a really good list with citations. If you’re doing problem solving and need data on what’s going on, ChatGPT’s new search option is really good too.
One last thing I’ll show you in Claude: I put together some financials — please graph these financials using React, which is just a specific graphing language. I had to mess with it a little, but it’s creating the code now — this is the code for what the graph looks like, but when it’s done it will actually produce it, so your data analytics and visualization — look at that, it shows financial performance and it’s interactive, you can get the exact number by clicking on anything. If you’re making presentations and you’ve got data to analyze as part of problem solving, which is very common in business — we see a problem because of the data, and we need to figure out what to do — you’d start with some data analysis, throw the data in, have Claude or ChatGPT help find issues and trends, and then visualize it. Super powerful. I could copy this graph and put it into the presentation too.
That’s enough of the magic — it’s not end-to-end automation, but there are plenty of pieces that were automated in there, and you got to see how we could accelerate the process, how we could amplify it by making it higher quality, and how we could augment it by bringing in decision-making frameworks and belief-shifting frameworks. I love that, I think it’s fantastic, and I wanted to ask: when you were running the prompt and kept telling it to proceed, I’m assuming the prompt already had your A4 AI integration method built into it? Yes — over here in Claude I was using what’s similar to GPTs but in Claude they’re called projects. The difference between a GPT on ChatGPT and a project in Claude is that a GPT is public, anyone can get to it, while Claude projects are always private to you. I put in the goal of what we’re trying to accomplish today and I have the four stages of the AI framework right here, so it knew to move from one stage to the next. I had it help me produce a list of instructions — one, two, three, four — and that’s how I was able to just tell it to proceed.
This is a great piece of not only acceleration and amplification but even automation, because if you’re going to do something over and over, you get a prompt that works, or a series of step-by-step prompts, and put it in a project and use it over and over — that’s exactly what I do. I’ve got a handful of prompts here for building email nurture sequences, building landing pages, and if you’re creating content to support a lead magnet or offering, you need social media content and videos — here’s content marketing pillars and posts. If you want to do events, I’d do one for micro-courses — Kelly, I know that’s something you love to do — I’d create a project for micro-courses. It’s a great way to semi-automate your process.
Thank you for that amazing presentation — there were hacks within your hack where I was like, oh, I have to do that, because even with Gamma, which is amazing, sometimes there’s a lot of formatting — I notice mine, Gamma likes to add content, and I’m assuming that’s because I don’t use markdown, so it’s probably more solid and follows the instructions more closely when it’s in markdown format, whereas I usually just throw text in the box. There was a setting I went over when you create new — I don’t know if I got the right text, let’s go back here — there’s a setting for how much content you want it to generate within Gamma. Yes, it was “preserve” — was it this one? Yeah, so if I just copied this and went to Gamma and pasted the content in, the magic here is: next slide, see this — generate, condense your content, or preserve. Since I did all the work in Claude to get it exactly the way I wanted, I don’t want it to change it, because I had that problem before where it was rewriting content I had already made with ChatGPT. So the key is to use “preserve” and it takes your content exactly as is. Here’s where you choose different image models — Flux is really good, like Midjourney for faces and stuff, and Ideogram is good when you want to include text in your images. And of course, don’t forget to tell it how many cards to create — each slide is essentially a card.
Thank you, amazing, and I can’t believe we did all of that in just shy of 20 minutes to have that presentation created and almost ready to go, aside from maybe looking at a couple of different images. I know you’ve been working with clients and helping them with their AI workflows, and I know this A4 AI integration model is newer, so I know you were doing the work prior to that. I want to ask, and maybe end with this question: in terms of thinking about automation at the end, how do you think that’s uplevel the impact you’re able to have on your clients with their own workflows, by taking this different perspective, where everybody else says let’s automate first — how has that impacted the results your clients have been able to see?
I think it’s much easier because they can feel a step-by-step process — you don’t start playing the piano by playing Rachmaninoff, that’s crazy advanced, and that’s where automation really comes in. You first have to understand the basics of AI, what this thing is, what its strengths and weaknesses and limitations are, how it kind of works, so you get a feel for it. Then you learn some prompting skills, then some advanced prompting skills, and once you get used to prompting, all of a sudden — oh, I can bring in a framework and build that into a multi-step prompt — it’s like, oh my God, now I can do anything if I find the right framework and learn how to write a good multi-step prompt and give it examples. That’s half of your automation right there, because you’ve got it all, so you can reuse it over and over with a Claude project or a ChatGPT GPT. Then if you want to push it out to social media channels and do some Make automations and other things, great, but the core really starts at the very beginning — you have to learn to walk before you run.
That builds confidence, and I’ve seen that with clients — once they learn some of this, some of them first have to see, oh wow, I can really do that, because they’ve tried prompts and gotten generic stuff out and gotten frustrated. But you show them an advanced prompt and what they can get out of it, and they become excited and willing to go back and learn the steps to build their skills, and eventually get to automation. It takes the pressure off, because they don’t have to feel like they have to automate everything in the beginning — they can still work from where they are, do it step by step, get a little better, and eventually automate. Nobody’s automating everything right now, they’re just learning how to do what we’re doing, so it takes the pressure off.
I love that, and I think you’re right — I definitely hear people who don’t even know what they’re trying to automate, or they bring people in like you and me and we automate things, and they’re like, I don’t even know how it works. I love that with your clients it doesn’t just help them integrate and leverage AI, they actually have ownership and understanding of what’s going on, which empowers them and allows them to still be well informed about what’s happening inside their business, instead of that knowledge being taken away by starting with automation first. They’re able to process and see where the automation came from, instead of just some guy coming in and building it for them. I don’t see a lot of agencies doing AI integration trying to preserve that.
Yeah, that’s part of my mission — to help business professionals who are tech-interested but not tech-savvy, give them a foundation to work with, some basic understanding, help them develop an AI mindset — we both had to develop this mindset about how we use AI — and then some basic skills they can build on, so they feel empowered to start using it and can have better conversations with people who are way ahead of them. If you don’t understand the basics, it’s hard to even understand what they’re talking about, so let’s empower the mass group of business professionals and educators to be able to do this.
I love that so much, thanks Bruce. I knew this was going to be a real treat having you on one of our hack sessions, and I think you more than met the goal — I saw a number of things people can go out and test, implement, and integrate right away, so thank you so much for being here with us today. Awesome, yeah, and if you want to know more, just go to my website, bruce.com, and direct message me, I’d love to hear from any of you. Amazing, thank you again for joining us, and thanks everybody for tuning in to this Scale with AI Summit hack session. I know a lot of you are going to be Googling Gamma if you didn’t already know about that tool, it’s one of our favorites, and you can see how powerful it can be when used alongside this A4 AI integration process. So I hope you enjoy, everybody, and we’ll see you inside the next session — take care.