People-First Governance That Keeps AI On Track

Big Question And Guest Welcome

Is there a potential resolution to this country? Well, my guess is yes, yes, there is. She wants to adjust the layer underneath this matter. That big workflow that can resolve this issue. The governance process is frequently the strongest story a brand has, yet almost nobody pitches it. Today, joining us from New York City is my guest, Katherine Tuominen, the founder of Crystal Brand Strategy. Her firm has been named the PR and social marketing partner of the year, USA 2026. Katherine, welcome to the podcast.

Thanks so much for having me. I'm excited to discuss everything today.

The Human AI Human Loop

You suggest that company workflow, the People First Governance Strategy, and AI can peacefully coexist. So how do you propose these these two functions can coexist?

Absolutely. Well, without getting too bogged down into the logistics for those of the audience who may not be tech pros or very familiar with AI, but the short version is I believe that these two can exist by using AI in the middle. So when we think about any process, usually it involves a human touch point at the start, whether it is prompting the AI bots for the first time or writing the governance reports or coming up with the frameworks and structures and then feeding that into the AI systems for that it to then generate, check, research, or inform, consolidate data, et cetera. And then it feeds it back to the human as that third and third touch point to then assess the quality of what was produced and to cross-compare with the original governance and of course use industry's best practice and your expertise. So from a very high-level overview, this is what we see work time and time again when it comes to whether it be PR infrastructure, marketing, or ensuring that the correct governance is applied throughout each of the businesses we represent.

Now, can we sort of go into some detail and maybe be more have a bit more specificity in terms of how we're going to what sort of programs can we manage to work on where we can make those kind of alterations or adjustments as you just described?

Absolutely. Well, we are platform agnostic, meaning we've worked across many different platforms. And of course, with old software, it's constantly evolving and updating. So what we may suggest today could indeed shift tomorrow.

Start With A Core Memory File

But across all of the new AI tools, they all use uh something called MD files or a core memory file. And this is really where you want to start when it comes to the agentic or AI journey. You want to make sure that you have a clear file that outlines the governance for your organization or representation. So this is where you would want to look at things like brand voice, tone, things that you can say, things that you can't say, and most especially governance laws and policy changes. And you want to make sure that in this raw file you have this clearly outlined so it knows exactly what the guardrails are and what it can and cannot do, both from a voice or language and narrative perception, and also even from an action and implementation perspective. So you want to look at this really end-to-end and make that document as fleshed out as it can be, while of course maintaining a low level of conciseness. So that way it can be easily read and picked up every single time you do an AI job or request something, it will always go back to that core MD and refer against the governance that you've put in place. So that would be my number one suggestion for someone who's starting out to use these tools, whether it's code cowork, ChatGPT, you name

One Rulebook Plus Department Skills

it.

Well, I guess my my question now begins when you start talking about workers doing this, what level of workers are we talking about? And are there different instructions for different departments so that basically when I say everybody's on the same page, but to sort of get us the outcomes that they want from the individual departments. Is this set up so that there's not a one-size-fits-all, or are these manuals of written information adjusted for different departments?

Yeah, yes, and so it's definitely both. You want to have a one-size-fits-all that applies to the entire organization. This goes down to, you know, brand values, voice, things that don't budge no matter what department it is being represented across, and of course, specific governance rules that are embedded into your, say, business plans and like legal setups. So all of that should be baked into the core document. And then on top of that, for those who are using tools like Claude Cowork or even agents across different software tools, you want to make sure that each department then has extra, they call them skills, but basically extra skills that then have their own, if if you will, sub-documents. So if you think of it a bit like a tree, you have the main trunk, the main stem, which is where everything comes from. And then there's branches for each department, which might be a specific skill for that department, which is a second document or a third document that then emphasizes like a clause, a process, an ops, or a way that department prefers to implement things, and that becomes the core framework.

Okay. All right. How would their structure look to the people that's in marketing so that they can do what they need to do and make everything coalesce and work together so that we get the outcome we're trying to get?

Absolutely.

Marketing Alerts And Content Repurposing

So for marketing, obviously, this is definitely one that we're very familiar with. So what we would recommend is firstly you can set up, going back to the governance piece, you can set up an alert or a scraping tool that runs every single Monday looking for whether it be the latest news cycles, a competitor that you're keeping an eye out on, or a policy change or government change. So we do all three of these, and it will give you that morning report, kind of like your morning newspaper, and it'll say, see the changes that have happened in the past seven days that you need to be mindful of. And so this will help inform what's to come for that week. You can also run this once a month to add like an end-of-month report or something along those lines, but that's a good starting point. So you're up to date with all the changes that are happening in such an agile industry. Then, from an infrastructural perspective, what you want to do is have a core hero message. So this would be, say, for example, that skill, it could be brand voice. And so this skill basically enables your company to have a cohesive message across all of the different subassets, if you will. And what we usually recommend for companies or brands is to have one platform that they prefer to focus on and then repurpose across the rest. So how this breaks down to from an operational or AI lens is say, for example, podcasting is your number one platform where you want the brand to become more well known. You can then take that podcast audio and turn it into a transcript. And that transcript can then become a newsletter that you send out, say, once a month or bi-weekly. And that transcript can also become at least once a week a blog post. So you can pull out kind of key elements and it flesh that out into a blog post. And then you can also take that transcript and turn it into captions for social media posts on other platforms. So little quotes, little snippets that can then become these, whether it be LinkedIn or Instagram or Meta posts. And lastly, if it was a video podcast, you could also clip that video and repurpose it into Instagram, TikTok, you name it. And so you can see that you can see how the system is broken down with just you and I talking. And then once you have a clear flow of what that end-to-end marketing or content ecosystem is like, you can then turn this into a document that is then fed into the box. So it knows exactly, okay, step one, I I research. Step two, I get the transcript from the latest podcast. Step three, I turn that transcript into the following different assets. And then step four might be to have a human eye to check it. I strongly recommend. And then lastly, to distribute it across the platforms. So that's a very simplified breakdown, but hopefully that will give your audience some ideas.

Cadence By Industry And Daily Impact

As I listen to this, I'm just wondering how frequent this has to happen. Because in marketing, let's say maybe you're only marketing one or two products, how much variation is there going to be from this week to the end of the month or something along those lines? Sort of help me with that because as as I listen to you, the variation of these changes might not take place that often.

Well, I think it depends on the the business type a lot. So, for example, if you're an e-commerce business, you would need to be rolling out changes quite regularly. Same as if you're in the technology industry, that could be a daily update or you know, daily things to to discuss or share across marketing. If it's something more at a steadier pace, like the education industry or government, then yeah, that would be a steadier pace. And you would just want to match the pace or cadence according to the level of frequency you have. So it may not be weekly or daily, it could be bi-weekly. And that goes back to the core MD or memory file that we talked about in establishing the governance and pacing.

Okay. So now help me with this because now we have the pacing at least for the individuals, because they have read this information, they've extracted whatever they need to extract for their marketing information or outreach for that particular week. How does this impact or how does AI and how is that impacted with what they're doing or about to do?

In terms of day-to-day management?

Right.

Absolutely. I think it's it impacts almost every aspect of a day-to-day role. I mean, you can think of it in terms of how long it would have taken you to research something in the past. At least I can speak for myself here. You know, when we're pitching to journalists or getting clients covered on big news channels, you have to know the latest news stories. You have to know what that journalist specializes in. You need to research what was trending that week. Now, all that research, which typically would have taken me between four to six hours, can take 30 minutes with these AI tools. And it gives you a summary and it checks it, and you can just read through it and come to that human judgment on what this may mean. So that's just one example. It's great for research. Another example would be for processes or setting up, I guess, like systems to check and submit. So, you know, whether it's checking your copywriting or checking your final product or the thing that you're submitting, you can feed that with a best practice example and say, have I missed something here? Or is all of this fact-checked and accurate? And it will go out and look at some of those details, which human error you might have missed the first time around. So there are so many different use cases depending on the department, but these are some of the ways that we're seeing it can save a lot of time and produce much more accurate results with that human eye to do the final checks, of course. Another example that may be relevant to your audience is when it comes to more of the manual scheduling. So, for example, in the past, we used to have to copy and paste things across or to tweak templates, or, you know, if you're using an Excel spreadsheet, for example, you'd have to fix the formatting. All of the sort of technical fixes and tweaks can be automated and fixed with AI now. So it means that you're not getting bogged down with some of the more fiddly details, and instead you can focus more on big picture thinking, critical thinking, and ensuring that you're moving more meaningful change within the organization.

So this is really not as complicated as let's let's say it might have originally sounded in the beginning. You know what it is you might want to uh or go about accomplishing, or looking at what your competitors are doing. Once you've done that, then you can make these adjustments in different departments. And that might only happen once a week or once a month. It's is that close to being accurate?

Exactly, yes. And once you've said it, it knows. So you don't necessarily have to keep re-feeding it. It's it's memorized it and it knows what that baseline is. Yes, of course, you can make changes based on what the feedback is, and you can also get it to self-learn and and improve, but the base is there. So you're really building an infrastructure that continued to work.

Board Oversight Without Blocking Work

So some of this information that that we're talking about now, just making adjustments, really is not impacted by but the by the board because that's not what they do. The board is probably more interested is more interested in larger issues of surrounding the uh the organization. So the they might uh still might make the adjustments, but uh when we talk about governance, that's uh boards probably make what once a month or something along those lines. Uh if they make some make some changes, then uh the AI might have to be adjusted so that those changes are caught and recognized and and and and uh and you know put into whatever documents or uh marketing materials that uh will be distributed. So it's a it's a timing issue on this as well, isn't it?

I see what you're saying, yes. Well, in terms of in terms of the timing, one of the ways that you can, I guess, protect it from continuously making mistakes for the full four weeks until the board meets again is in that initial governance, is making sure that there's enough wiggle room so that the documents aren't so strict that you know team members, employees, etc. can't even move without getting a board's approval. It needs to have enough room so that they can continue to do their day-to-day work and use that to inform what could be improved. And so that would be a conversation that board members can review internally first, pre-approve that governance, and then it should go out with enough room so that people aren't essentially waiting four weeks for the board to meet again.

Pardon me, I just wanted to make sure that we're not making people think that there's some oversight by the board, because in most cases that's not what they do. They're more big picture people rather than you know what happened this week as compared to what happened that week. Yes, they might get involved if those things are continuously happening at the big snap or explosion someplace, and they might want to know what the hell happened here. So it's it's that the board does not get involved in the day-to-day activities of any major data because they're talking at a different level. So most of them they're not looking down at the day uh the daily activities. They just want to know what's happened this month or that month. And hopefully the uh the the troops, if you will, have made the adjustments because they have recognized the changes that need to be made uh right then and there.

Exactly, yes. So this means that it we really functions cohesively, and by the time it sort of makes it up to the board, it would be more about doing that final reporting, like you said, the high-level important changes that have come across rather than the day-to-day.

So what understanding here that the board would be interested in is that you know they fine-tuned how they go about making their marketing approaches or outreaches or other things because they're getting information coming in either daily or weekly, however it's set up. And then they can make adjustments so that things can continue to uh move smoothly and they can make any adjustments to it and also outside circumstances that they may not have anticipated. That's the goal the board would like to hear about at their at their monthly meetings.

Exactly. And one of the ways this could be useful is also by thinking of it in terms of loops, in the sense that you can get you can use these skills and these MD memory files to self-iterate and learn. So if you if you get it to run a loop, for example, it can run a daily root loop or it could run a leak weekly loop where where at the end of that week it assesses what broke, what didn't work as well, what could be improved, and it ranks it against itself or like the best case benchmark that you have. And then then it would give you this sort of report, and then next week it would it would improve itself and feed. And so what this means is if the board weeks meets on a monthly basis, you essentially have four rounds of improvements to present to the board to say this is this is how it's improved and this is how we're continuing to see

Catherine’s Path Into AI

progress.

So let me ask you, how did you become involved in this particular circumstance and using AI and looking at how it can help, let's say, the troops, the ground troops, if you will, improve their performances. How did you become involved there?

Well, I've actually been involved in AI for a long time, possibly before it's before it sort of reached what it is today. It was used a lot in the early, earlier stages of development for research and almost like light copywriting frameworks. So it wouldn't necessarily write for you, but it could use, it could kind of say, here's a headline idea, or here's a main body context. And so because everything we do comes back to copywriting in some form or another, we would use this to create to inform structure and also to do the research piece that I mentioned. Now, of course, it's so much smarter and you know, a lot even able to able to go even deeper than before. But that's really what began my journey. I got very curious about, you know, what are the latest tools out there, what's available, and how can we really make sure that we're being efficient and staying up to date with all these changes because we work in such heavy compliant industries, and it helped me recognize the importance of staying up to date with these changes. So that was really my the beginning of my journey. I I'd say close to four years now of starting to use AI and figure out ways that it can be used as a tool to strengthen our output.

Which Tools Fit Which Jobs

What sort of AI platforms do you use to perform individual activities? Because the, you know, I think we know that no one AI platform solves all these problems. Is there one that you use specifically to achieve, collect specific types of information for this particular industry?

Yes, that's a great question. So perplexity is very good when it comes to deep research. It's kind of known for being the research platform. I like to think of AI as different locations in a city. So perplexity would be the library. Gemini might be more like the cool cafe or a place where people can chat and create. And likewise, we also love using Claude Co-Work, which I would imagine to be almost like a government office or something like this. It's very kind of formal and official, and it values critical thinking and governance almost naturally. So between perplexity and cloud co-work, we love using these two for our kind of day-to-day management. And then we also use tools like GitHub and NADN for more complex agentic codes.

Well, that's interesting. Yeah, I I simply asked that question because I use AI a lot myself. It will produce a uh a transcript of our conversation today. And basically, all I have to do is put uh when it says speaker one, I put my name in speaker two, I'll put your name in there. So it'll make that available to uh to listeners. But then it will also conduct what we've talked about and give a brief two or three paragraph description of what we talked about. So that helps people get a better idea as well. So we used AI. Suggested that I like, but not completely, and I'll make some adjustments and go with that. So uh AI, I mean, certainly has its uh finger and a stronghold on darn everything. And uh exactly and it makes m what I do here a whole lot more move faster and quicker and more efficiently, and um certainly other people look at it the very same way.

Exactly. Yes, we like to think of it as a thought partner. Um it doesn't do your thinking for you, but it can uh give you some ideas and speed up the process, that's for sure.

Guardrails For Access And Email

Are there certain things that when people are using AI that you find out that they need to look be careful about?

Absolutely. So you don't ever, for example, if you're using quote cod or any of the tools, they give you the option to have full access to everything. And you need to be very careful about what you're giving it access to. So, you know, that's not to say it shouldn't have access, but if you are giving it access, you need to then create guardrails around what it can and can't do within those access. So to give you an example, inbox is a common one. Reaching inbox zero is a dream that many of us may have given up on, or at the very least, set aside. But with tools like AI, it becomes so much easier to measure inbox because it can organize things for you. It can draft responses, it can flag things like you said, for a morning report to kind of say this is what you have coming up for today. So in our case, we allow it to read and we allow it to draft, but it is not under any circumstance allowed to send. And so this is a guardrail we have set up in place so we can keep up to date with all the latest pitches we have coming through and the client representation and opportunities without necessarily ever risking accidentally sending something against our approval.

Well, Catherine, you presented us with a lot of information today. Is there anything that you think we may have missed that uh you should let us know about in this AI issue and you know getting information out uh uh from our organizations?

Build Repeatable Systems With An Audit

Absolutely. Well, there's so many different use cases as you've sort of touched on. I would say one of the biggest things that people should, I guess, think about is try not to use AI as just a like a question and answer tool where you might chat chat to it or ask it something one time. What you really want to do is create repeatable systems. So even before you start using AI, you may want to do an audit of your day-to-day and think about where that time in your day was spent. You know, did you spend 30 minutes reviewing something? Did you spend 30 minutes researching something or going through your emails, whatever it may be? You can kind of audit your day to figure out what are the lowest value tasks that I've just been doing because you know, we've always been doing it, and what could you be feeding into AI instead from that? And you can even use these tools to kind of say, I want to save, you know, four hours per week. What would you take off my plate or what can I automate using these like skills and and auto automated loops? And it will come up with a plan for you that will be customized to your day-to-day management.

Wrap Up And Final Takeaways

Well, let me say that uh I've enjoyed this. I mean, you certainly enlightened me about that, and I'm certainly going to look for other ways that I can use AI in what I do right here at this at this podcast. So let me say I want to say That's excellent.

Really glad to hear it. Hopefully the audience thinks so as well.

Well, you know, that's one reason I've asked you to come on because I was certainly interested in, you know, when I see things that make sense to me from a from a PR point of view and a podcast point of view, I I certainly want to talk to people about that. And it so happens that there are probably a lot of listeners out there that want to hear this as well. So that's why I want to say thank you for reaching out and the fact that uh we've had a chance to talk and discuss this issue, I think is going to bring a lot of insight and uh a better understanding to uh to our listeners.

Absolutely. I hope everyone gets value from it. I appreciate you, your your your time and all the great questions you've been asking.

Well, let me say, Catherine, thank you so very, very much for being a guest on the uh Public Relations Review Podcast today.

Reviews Sharing And Next Episode

And to my guests, my listeners, I hope that uh you will have learned a lot and can use some of this information on what you do in your day-to-day w work. And also, we'd like to also get a review from you and uh share this with your friends. And don't forget to listen to the next edition of the Public Relations Review Podcast. Have a good one. Bye.

Thank you for joining us.