Mitchell Feldman - Prompt Insights
- John Ennis
- Mar 27
- 21 min read
Updated: Apr 8
Welcome to "AigoraCast", conversations with industry experts on how new technologies are transforming sensory and consumer science!
AigoraCast is available on Apple Podcasts, Spotify, Podcast Republic, Pandora, and Amazon Music. Remember to subscribe, and please leave a positive review if you like what you hear!
Mitchell Feldman is a seasoned entrepreneur with over three decades at the forefront of technology and marketing innovation. As CEO of Jam 7, he pioneers the fusion of human creativity with Agentic AI, transforming the landscape of B2B marketing. Mitchell's vision has established Jam 7 as a leader in AI-driven growth strategies, empowering ambitious tech brands through intelligent, autonomous Growth Agents that amplify human expertise.
Mitchell's career highlights include being named the first-ever Microsoft Worldwide Cloud Partner of the Year, winning the UK Technology Entrepreneur of the Year award, and successfully leading a strategic acquisition by Hewlett-Packard Enterprise (HPE) in 2018. Passionate about the synergy between people and technology, Mitchell continues to redefine the future of marketing by championing collaborative frameworks where AI and human strategists work seamlessly together to deliver extraordinary business growth.
Transcript (Semi-automated, forgive typos!)
John: Welcome to the show, Mitchell.
Mitchell: Thanks for having me, John.
John: Okay, great. Now, your field is somewhat adjacent to ours in sensory consumer science. So I think it'd be good if we could start by you going through a little bit about your background, and then we can talk about why we're together on the show today.
Mitchell: Sure. Okay. I'll say that without revealing how old I am, in the early '80s, I was coding in basic and became a tech enthusiast from very, very early days. My career has always been in technology, typically in B2B technology. As you kindly mentioned before, I had won the Microsoft Worldwide Cloud Partner of the Year and I exited my business to Hewlett-Packard enterprise to build up their cloud offering globally. I've always been fascinated with marketing. One of my favorite expressions is I've written books on marketing, and checkbooks, because I've got it wrong lots of times, and I've spent lots of money getting it wrong. I then realized I had that. I made enough mistakes that I eventually found what I thought was the right formula for marketing. And that was where the intersection of my tech background and marketing background collided, and then it just mushroomed to becoming what was a hobby, to becoming my career, to now becoming my sole job.
John: Yeah, it's very interesting because we connected on LinkedIn because I noticed that what you were doing in the marketing space was very similar to some of the ideas I had around what we need to do at Aigora. Aigora is a consulting company that became a services company. I think with AI coming along, I think consultants really need to think about how they can continue to provide value beyond what people can get just from LLMs. I think you've got some really interesting ideas on this topic. Maybe just open the floor for you and talk about the changes you're seeing and your vision of Jam 7 for how consultants can work together with AI to provide value maybe more widely than they originally and how you see AI and consulting evolving together.
Mitchell: Funny junction at the moment where if we look at the hype curve, I feel that we're still at this trough of disillusionment where people don't quite know. They feel like they're sitting on a beach where they're aware that there's this breeze. I don't think they're quite aware that that breeze will eventually turn into a tsunami, and that tsunami is AI. The world is changing. I think what people have still not quite got their head around is that AI is not a replacement for humans. It's actually a tooling. It's another companion. It's another work colleague. But I think the world is radically going to change. It's radically going to change. There's stuff that LLMs or AI as a whole can do that humans cannot do. I'm certainly not at the scale and speed that the LLMs can work at. But for me, what's changed in my world is the speed of execution, getting to the source of the truth and an understanding of a subject matter in more depth and more concise than I've ever been able to do before. I mean, deep research It's a game changer, even if you're... You're looking at... We're a marketing company, and we find that a lot of companies have great products and great people, but they've never really established product-market fit. So using things like deep research tools to really validate the value proposition and to understand the pains, gains, and jobs to be done on the people that you're selling to, to get to that information, I don't have to hire a team of analysts anymore. It's just a command line prompt into the CLI of my large language model, and I've got that information.
John: Yeah. Well, maybe you'll be good to take a step back and talk about what you guys are doing at Jam 7, the overall project and the services you're providing. Can you just take us through what you're doing there?
Mitchell: Yeah, with pleasure. Thank you. So the one thing that I think a lot of people have a problem with marketing is consistency. I and measurability and speed. We classically say when we're trying to describe our service that you've got the iron triangle where you've got speed, quality, and cost, and you've always had to give up one of those to the other two. Actually, I saw with AI is actually I can have all three. I can have speed, quality, and cost. What we've done is that we've built a marketing platform, which we call AMP for the agentic marketing platform, which is a series of large language models, each with a discrete set of tasks that perform the end-to-end requirements for marketing for a company. And that could be through identifying your ICP, your personas, to understand your target market, to creating the content, to creating the strategy, to creating the analytics, to then be to learn from that data and in real-time self-heal so that they become performant. So I think, broadly speaking, we go into a company, we extract all of their DNA, how they talk about themselves, why they talk about themselves, why they exist. We train the large language model on all of those data sets. We then finally tune it, and then we present our clients with an agent which is able to produce anything from content to strategy to continuous service improvement.
John: How are you collecting the data then? When you're doing this initial research upfront, is this through chat bots that are disseminated throughout the company?
Mitchell: Yeah, exactly. We have, as I said before, this agenda marketing platform, which effectively is a direct to level agent, and underneath those, all of these discrete agents. One might be a copywriter, one might be a researcher, one might be a brand police, one might be a creative designer. And if I look through the lens of just the researcher, the researcher is going to look at the company, it's going to look at their competitors. So it's doing a whole of web search, looking at the competitors, looking at the market opportunity, listening to social listening on tools like Reddit and X, to listen to conversations to validate all of the things that the customer has told us, to see if actually what they're saying is correct product. Actually, a lot of smaller businesses suffer from, they think they're the man in the mirror, and actually the customer is asking for different things. So really, we're using it to understand the language of the buyer. There's a great book by Marcus Sheridan called They Ask, You Answer. And it's using marketing to actually answer the questions of the customers rather than telling everyone about your product. And In fact, we even broke down that they ask you answer to an acronym because we're in technology and everything has to be a three and four letter acronym for FLA and TLA. And we called one of our agents, Teya, for that very reason, they ask you answer.
John: Okay. I personally can't wait. Honestly, I cannot wait to be a customer of yours. I really feel like... Because at Aigora, we had a bit of the opposite problem, where we have product-market fit, but we have had basically no market. And It's just because of the strength of my network and the network of the other consultants that we've been able to get business. In the early days, I put a lot of work into discovery calls, trying to figure out what do people need. But I feel like, this sounds like exactly what we need to go through this process. It was always one of these things where we're small business, and so I couldn't really justify hiring a marketing agency. I knew that we needed it. I also was worried about how much time is this going to take, just all these details. And it just sounds like a perfect thing for companies like mine to work with. And I think that if I can just talk a little bit about what we're doing at Aigora, and you can respond to that because I think maybe it'll be interesting to you. I think that chatbots open up all sorts of doors when it comes to collecting information from people and helping groups to collaborate. That we see that all the time at Aigora now. When we go to make proposals, it used to be consultant would have a call, maybe client would fill out some survey, and then the consultant would make the proposal, and they might send it to the tech team. But there was always, the projects were never really properly scoped because there are all these technical things that consultants not thinking about, whatever. But now when we make proposals, what we do is it's just starting with a chatbot from day one and describing the project. We have vast examples. Then we actually get the chatbot to say, What are the technical questions we need to answer to really propose this properly? Then it goes to the technical team. Now it is this tremendously collaborative experience where we have a shared document, tech team is working on it, we're bringing it back to the AI all the time, AI is processing it, what questions do we still need? And we end up with a much, much tighter proposal. I'm wondering on the marketing side, what you're seeing that's analogous to that, where you have LLMs working together or LLMs helping people to work together to be more unified in their..
Mitchell: Yeah, I totally get it. And for me, the secret source in marketing today is hyper personalization. And you now have the ability to do that, because imagine if there was another layer on top of what you're doing when you're doing these proposals, whereby you've had the conversation with a client, you've had that chemistry meeting where you've counted each other's fingers and toes, and you realize there's It's a nice chemistry there and you could do stuff together, and then you're doing the proposal, and you've got all your notes that you've made, whether it's your AI note taker, or you've been doing it humanly or writing even on your remarkable. But imagine if then there was an agent that said, Thank you for the proposal, John. Great proposal. I'm now going to look at the client's social footprint and understand how they talk and what language they appreciate, and then using that personalization layer through the AI to talk to that customer in their language. That's never been possible before, and you know as well as I do, that's a prompt. A little bit of engineering and a prompt. Now, not only do I have this beautifully crafted, bespoke proposal, but it talks in the style that my customer wants, how much detail do they want. What's the vernacular that they use when they're talking? That just changes the game. I think, if anything, I read lots of cases online about AI, how it's making people dumber. Because they're not having to think anymore. Actually, the smart play is to start thinking beyond what was possible before, because forever and a day, people have been doing that engagement framework where you're doing a quote for a customer after you've had that chemistry meeting. But to now bring in that additional modality of how they talk online into that proposal, that's just someone's idea. Now, the AI is not going to come up with that idea. It will execute it, but it still needs that human expertise. That's what we say as a company is that AI is not the silver bullet. It requires human expertise expertise fused with AI because it needs to have that human in the loop. The one thing that AI is not very good at today, and it's good at lots of things, is personality and creativity. So that world will exist for quite some time.
John: Yeah. I see that for sure. What I'm finding, the more I work with AI, and we're using AI all over the place now at Aigora, I mean, not just for analysis, not just for chatbot front ends. But for example, project management. We're using it for project management. We're using it everywhere. There's always another level. And I mean, you have experience in the early computer programming world. I mean, once upon a time, it was good assembly. You had machine language, and then you have compiled language. You keep going up the ladder. Then you've got, you have scripting languages. And you have more and more advanced language frameworks that operate at a lower level. And now, over the weekend, my other company, Tulle, works with a game on Solana and someone discovered, you can ask Grok for details about the game, and it does a pretty good job. And so people were asking me, Okay, well, what can we do with this information? I said, well, we need trusted sources. So I asked Grok for a list of its sources, went through, and said, Give me a template for prompts. Then I went to Gemini and I said, All right, look, I want instructions for a prompt maker to make prompts for Grok to find out about this game. And you just keep going up levels. And there's always new levels. And the pyramid just keeps getting bigger, and you just end up higher on the pyramid. So I personally don't believe that AI is going to put us all out of jobs. I think it's just ideally going to make us five times more productive and twice the work. And that's the 10X.
Mitchell: I mean, everyone that I come across, try not to sound too evangelical about the AI, is the best gift you can give to your future self is to do a prompt engineering course.
John: Oh, do a course? That's interesting.
Mitchell: I encourage everyone that I meet. There's a company I met online called Coursera. They do course ERA, and they do a prompt engineering course. I think it may be free, it may be $99. It's very snackable bits of content, and it teaches you the basic fundamentals of prompt engineering. Anyone that I've said on that journey, I've almost lit the touch paper because all of a sudden, their creativity starts blossoming, and they start looking at the world through a different lens. That's what AI has given to them is the ability to think in a different way and address age-old problems with new technologies and new know-hows and capabilities.
John: That's fascinating. I've been reading articles on prompting, and I've been... I very often ask LLMs to help me craft prompts, and then I get the prompt. In fact, I have machines that make prompts. That's interesting. It is true what you said. Everything is... Once you start thinking in the terms of prompts, you just see them everywhere. Everything's a prompt.
Mitchell: And literally, I look through the world, and I've always had this, that I look at the world through ones and zeros. I walk into a restaurant and I'm like, Why have I got a paper menu in front of me? Why isn't this digital? Why can't I pay for my meal at the end by putting my NFC card onto the table. So I'm looking at the world thinking, it's never been more pervasive now. Literally everything I can sort out from my traffic light signal is not changing, even though there's no one coming the other way. I can write a prompt to resolve that issue.
John: You submit a little request to the light to change, please.
Mitchell: Exactly. So the world is changing, and it's changing dramatically. I come from the a lot of cloud, and we used to joke that the world was going at cloud speed because the level of engineering and innovation was happening at such a rapid pace. It was hard to keep on top of it. Well, if I thought cloud was fast, AI is even faster. You've got to really understand it because if you don't understand it, it will start eating your lunch. It will start eating your lunch. I mean, look, if I can just go down a little bit of a rabbit hole, I have a daughter who is neurodivergent and struggles at school. She has a teacher at school that is engaging with her where there has to be parity across the rest of the class. So the training is not meeting her on her journey. Well, actually, if I had an AI teacher, an avatar, that understood my daughter and her journey and where she was at and where she needed over indexing on to learn more and craft content, and craft content at scale. My daughter's 15 years old. She would much rather learn through a set of reels like in TikTok, because that's how she consumes information in these seven second bite-sized chunks. If I could create a world based on AI that knows the curricula, also understands the individual that's learning it, and in real-time, learns their journey, then what does that leave schools?
John: That's right. I'm not saving for my kids to go to college. They're five and nine. I don't think college is going to be.
Mitchell: I totally agree.
John: Yeah. Everything is just changing so rapidly. In fact, it's interesting even with the course, I'm learning more now. I'm learning at a more rapid rate now than I ever have in my life. Every day. I've learned chatbot engineering in the last six months, and every day I'm learning new things, new project, reporting to the Google tools, and I'm using dialog flow, and I'm just learning as I'm doing it because it's just constant tutorials.
Mitchell: The LLM is your greatest teacher you've ever had.
John: It is 100%.
Mitchell: Because you're just limited by your curiosity.
John: Yeah.
Mitchell: Everything is changing. Even search engines is changing.
John: Yeah, that's right.
Mitchell: Historically, you type in flowers near me, and it will give you a list. Actually, what the LLM will now say is, Right, I understand you want flowers. I've realized it's coming up to Valentine's Day. You're going to probably want roses. And these are the ones that have been best rated. Would you like me to place the order for you?
John: You have to change the way you think about it. I was actually making fun of the CTO of my first company or of my second company, Tulle, this morning because we were scheduling a meeting with some people in Turkey. It was going to be 6: 30 in Spain. What time would that be in Turkey? He was trying to figure out the time difference, and he was doing a search. I was like, Hamza, you're a CTO of an AI startup. Go to Grok and type in, if it's 6: 30 in Spain. What time will it be in Turkey? Just ask the question. It just totally changes the way you just... Because it's all about what do you want? Getting down to your goals. What are you trying to... What are your intentions? And then crafting prompts to accomplish your intentions. That's what I see.
Mitchell: I'm bringing you stuff you didn't know. I had a great example that we're doing some work for a fabulous SaaS business in the US, and a part of their market is in the Nordics. So we've built them this really great storytelling sales deck, and the metaphor that we use within this business is around mountaineering. And one of the slides was around the impossible choices of being between a rock and a hard place. And we said We're going to be presenting this as an event in Finland. Can you please rate my presentation? You know what it came up and said is that between a rock and a hard place is not language used in Finland, and it's between a tree and a bark. We had to change the presentation we would never have known. If we were to say between a rock and a hard place, people would have been like, I don't get that bit, and they would have maybe zoned out. Again, that hyper personalization and giving me information I didn't even know to ask that conversation or that question, it gave it to me.
John: Yeah, that's exactly right. That's actually one of the skills of prompting is being open and being careful not to say, how do I do it this way? But just generally speaking, how do I do this? What's the best way to do this? And being surprised sometimes by that.
Mitchell: And I find sometimes, and you've probably had it before, that you can get lost in this vacuum of having this conversation. I find sometimes it's better with these prompts that you can get to a form that you're relatively happy with and then move that output into a new conversation and start again. That's a really good tip, and it works really, really well.
John: I actually see that a lot in coding. Once you start to go down a dead-end in coding, it's good to say, "All right, how do I solve this problem?" Okay, and then you get a bunch of stuff. Some of it doesn't work. You get a bunch of errors. Sometimes it's better to start over and say, All right, how do I solve this problem? And watch out for these errors. And then off you go. I think that people need to be practicing using these tools as much as they possibly can in every aspect of their life. Last night, I was revamping my task management, like personal task management, using these tools. Use them for everything. But okay, let's talk a little bit about agents, because I know that's also a big part. We've got the recommendations with the LLMs. Can you talk about some of the agentic capabilities? We mentioned that at the beginning. How does that fit together with... You've got these front ends, which you can think of as consultants, and then you've got agents. Can you talk about that second half?
Mitchell: Well, so I can talk about what we do. Most users who are going to be listening to this podcast have used ChatGPT or Claude or Gemini or another, and they will use a single instance of that. The biggest problem that we have with using those traditional large language models is typically hallucination. That's the biggest problem that we have. We felt the only way that we could deal with that is to build out an agentic mesh where we have a series of agents with very discrete tasks, very, very fine tasks who then know the orchestration layer that we've designed is they know when to stop and start and who to hand it off to. So As I mentioned earlier, we've got this director level. He will get the request. He realizes that we need a researcher first. The researcher will do it, will then challenge that reasoning and validate that reasoning. Once we're happy with that and we've got to a high degree of confidence, that's right. That will then be passed over to the brand manager to make sure that we're using the right taxonomy, that we're using the right language. That will then pass over to the persona agent who is just trained on our personas to understand the pains, gains, and jobs to be done of our customers, to then redo the blog. We'll then have someone who will be an email agent who will know what are the high spam words and how to the best deliverable rates on these emails. Each of these agents have a very, very discrete task. What we've done is we brought in a series of experts into the business and we codified them. We are turning their skills, their knowledge into prompts. The way that I talk about the amp or the agentic marketing platform to my team is, I put it as a king to a black box on the plane. The reason where planes are the safest method of travel is because they have a black box and they understand everything that goes on, and they have rigorous checklists. If anything breaks or anything goes wrong, they look at the black box, they autocorrect it so it doesn't happen again, and they keep going until they get it perfected. It's exactly what we're doing with these agents. Every time we find an anomaly in one agent, we fix it. Sometimes it's a little bit like playing whack-a-mole where trying to hit each one down when they come up until we get this production line where we're getting the content, and the content is coming out typically like a... If we're average eight and a half out of 10, if we're doing really well, maybe a 9. 2, 9. 5 out of 10. And then it just needs that human in the loop to check out the content and distribute it as they want. So for us, an agenda MASH is a collection of agents with very discrete tasks that don't move out of those swim lanes. And by doing so, we can then ensure that we reduce, if not eradicate, all hallucination, which is the biggest problem when it comes to marketing.
John: Yeah, I know. That's really wise. Because you see that even... I mean, you don't want your email agent altering the content of what it's supposed to be sending out. It shouldn't have its own ideas about the content. That job is done. It's job is on the email.
Mitchell: Exactly. And this is what pretty much every company in the world hasn't solved yet, is to have that single source of truth as a brain that sits as the foundations for all content that comes out of an organization. We see it very often. I'm looking at it through marketing, but we can say sometimes marketing is underperformant, so sales will create their own version of the story because they feel that marketing aren't giving them what they want. Then it veers away from the truth because salespeople do that. Disclaimer, I was a salesperson. Solving that problem where you are harmonizing and synthesizing all of the data in the business and agreeing what is the correct source. Then letting the users ask the questions, but getting the answers that are aligned to the company's goals and values and principles is the biggest opportunity for every business.
John: That's fascinating. I definitely say that it's synthesizing information within these AI systems that you have information sources from all over and you want to synthesize them together. AI really excels at that.
Mitchell: Yeah, but again, you look at something like Copilot, which is a great product, and they're trying to scoop up everything that's in your email and your SharePoint, but you have to go through that process. This is garbage in, garbage out. Unless you've applied a level of hygiene to the training data, then it's always going to be wrong.
John: Now, that's right. People think that sometimes you have a lot of data, and so you average it and it'll work out. Well, there can be all sorts of biases in the data. If you've got biased data and you average it, you have biased average. Yeah, that's right. Data cleanliness really matters. It's been great. We're basically at time. I can't believe that we've flown through half an hour here. But do you have any advice? The people who'll be listening to this podcast will be typically sensory consumer scientists. People are working on products. They're going to be interacting with marketing, but they'll be wanting to use AI in their own lives. What advice do you have for people who are trying to figure out what's going on in the AI space? Maybe they're a little even scared that they're going to lose their jobs. How would you recommend that people make it through this transition that's coming right now?
Mitchell: I would say that the new skill is prompt engineering. It's to start to learn how to use the AI. We're seeing it, if you look through a clinician's lens, I noticed that a couple of weeks ago, Microsoft have announced their new quantum chip or their first version of the quantum chip, and it managed to solve or understand a disease that a team had been working on for 10 years, their quantum chips solved it in two days, and they still hadn't got to the answer. The clever person was not necessarily the person that built the chip, although they were clearly very clever, was the person that was asking that chip the question to get the answer. And that is prompt engineering.
John: Yeah, I totally agree with that. Yeah, I see prompting everywhere now. It's like intention, prompt, and then you get your result. How successful you are at getting what you want from the machine. That's really the spirit of prompt engineering.
Mitchell: Spot on.
John: Yeah, that's right. And I would advise people, use the LLMs to help you to prompt. Say, I'm trying to get this, but what's a good prompt? How can I... And have it if it's not clear. That's a big thing. People are sometimes not clear on what they want. And the LLM can really help you to figure out what you want.
Mitchell: That's a smart piece of advice, is getting the LLM to help you write the prompt will by proxy teach you how to use LLMs.
John: Yes, 100%. Okay, well, Mitchell, this has been a great conversation. Any final words? How can people learn more about Jam 7?
Mitchell: Jam 7, jam7.com, or hit me up on LinkedIn, ignore my Elon Musk post that brought us together, and they'll be able to find us if they can sift through the thousands of comments.
John: Yeah, it's interesting. Actually, it's interesting to think about LinkedIn posts as prompts, and you're prompting the LinkedIn hype mind. You're seeking reactions, and prompt engineers.
Mitchell: Well, imagine the world where the large language model has been highly trained on your business. It is then doing social listening to see what is trending online and then pairing that trending topic, bringing it to your business and then using that as material to start marketing so that you're being very clever, that you're riding on trends, but you're being clever enough to pair it to your personal business, but in the language of the people that are buying your product. That's smart.
John: Yeah. No, it's good. I honestly can't wait to use your services. So I'm sure our collaboration will continue because we need you. I think-
Mitchell: Thank you. It would be a pleasure.
John: Yeah. We're looking for a solution like yours. Okay, well, Mitchell, thank you very much.
Mitchell: Thank you for having me, John. It's great.
That's it. I hope you enjoyed this conversation. If you did, please help us grow our audience by telling your friend about AigoraCast and leaving us a positive review on iTunes. Thanks.
That's it for now. If you'd like to receive email updates from Aigora, including weekly video recaps of our blog activity, click on the button below to join our email list. Thanks for stopping by!
Comments