Brinda Gulati: Don't Summarize Transcripts With AI

Brinda Gulati is a B2B content writer and strategist. She turns interview transcripts into content that ranks, and her first rule surprises people: she does not feed the transcript into AI to summarize it.

Key Takeaways

Brinda reads transcripts by hand first, to catch phrases unique to the speaker.
She uses AI as a "rubber duck," asking it questions to find her own answers, not summaries.
Her ARS framework: make the language accessible, the emotion relatable, the example specific.
Refresh B2B content every six months. Just changing the date backfires.
The first two AI outputs look fine. Across a hundred, the quality breaks.

Why she won't feed the transcript to AI first

Brinda starts every transcript by hand, page by page, before she opens any AI tool. The reason is simple. AI smooths language into the middle of the road, and the middle of the road is where original thought goes to die.

When she reads it herself, she catches the things a summary throws out: a turn of phrase, a word choice, a way of framing a problem that is unique to that one speaker. Those are the gems. Once she has one, she takes it to AI as a thinking partner, not a writer.

"I use AI as a rubber duck, which is basically you keep asking it questions, it asks you questions back, and in that process, you find answers for yourself." — Brinda Gulati

So the order matters. Human first to find the gem, AI second to pressure-test it.

The fork in the road on AI content

Brinda sees two kinds of B2B brands right now, and they are splitting apart.

One group went all in and automated the entire content lifecycle. After about six months, many of them come back and say the same thing: actually, we would like a human in the loop. The other group kept a human and a strong narrative voice from the start. They are not panicking after every Google update or every new model release.

"Slop is bad, yes, but saving money is better, right? That's always the case." — Brinda Gulati

Here is the part most teams miss. The first AI output looks good. The second looks good. The problem is repetition. Across two outputs you can get something decent. Across a hundred, the quality does not hold. It breaks.

The ARS framework: accessible, relatable, specific

Brinda's framework for making content land is called ARS. It stands for accessible, relatable, and specific. You apply one of each, starting with a single keyword, then build from there.

Accessible: plain language, plus the jargon your buyer already knows. She will say "streamline processes" because that is familiar to that audience. Knowing when to use jargon is a judgment call, not a rule.
Relatable: write from something you have actually been through. She ran an Instagram thrift store, so she can speak to e-commerce from real experience.
Specific: ask the expert for one example from the last five years that had a measurable impact on their business or life.

She uses ARS on her own subject-matter-expert interviews too. Better, more targeted questions produce better answers, and those answers become the raw material a writer can actually use.

Why writing comes last

For Brinda, writing is the easy part. The hard part is the judgment that happens before a single sentence is written.

"It's a repository of judgment. So it's up to me what to exclude and what to include." — Brinda Gulati

That discernment comes from years of seeing what bad content looks like, so you can tell what good content could be. Give a senior strategist and a brand-new marketer the same transcript and the same AI tool, and the outputs will be wildly different. Blanket prompts like "remove all jargon" and "give me the most impactful summary" produce things that have already been said a hundred times. The thinking is the value. The writing is just the package.

How often to refresh your content

Brinda's answer on cadence is six months. She admits it is not budget-friendly, but stale content is worse than no content.

The shortcut she sees most often is companies updating the publish date to fake freshness. It backfires, because the old statistics, old examples, and outdated laws are still sitting in the post. You do not need to chase every product change in a separate change log. Your content does need to reflect the recent changes that actually matter to your buyers.

Her one tip: read fiction

When asked for final advice, Brinda did not say anything about tools. She said read fiction.

We optimize everything now, even our reading. Tracking how many marketing books you read per year turns reading into another metric and quietly kills original thought. Science fiction gives you variety in language, variety in thought, and new ways to think about the same problem. That variety is where fresh ideas come from.

Frequently Asked Questions

Should you summarize interview transcripts with AI?

No, not first. Brinda reads the full transcript by hand before touching AI, so she can catch phrases and word choices unique to the speaker. Those specifics get flattened when AI summarizes. She uses AI afterward as a thinking partner to test an idea, not to compress the source.

What is the ARS framework for content?

ARS stands for accessible, relatable, and specific. Make the language accessible, make the emotion relatable, and make the example specific, one of each to start. It gives you clarity on what to actually say to your buyer.

How often should B2B companies update their content?

Brinda recommends every six months. Simply changing the publish date to look fresh backfires, because old stats and outdated examples give it away. Update the content to reflect the recent changes that matter to your buyers.

Should B2B companies fully automate content creation with AI?

Brinda sees a fork in the road. Teams that fully automate often come back after six months wanting a human in the loop. The first couple of AI outputs look fine, but across a hundred the quality breaks down.

What is the best way to come up with original content ideas?

Read fiction. Brinda says optimizing every input, including your reading, quietly kills original thought. Science fiction gives you variety in language and new ways of thinking about the same problem.

Full Interview Transcript

Dane: Hey everybody, my name is Dane Frederiksen and I am a video expert in the B2B space and I'm on a mission to help companies get visible, trusted, and build pipeline faster with video. And today I am joined by the wonderful Brinda Gulati, who is a B2B content writer, strategist, among other things. Welcome, Brinda. What else do we need to know about you and what you can do?

Brinda: A lot probably, but I don't think we can condense that into this conversation. No, no. I have an unhealthy obsession with dogs. I got into content marketing about five or six years ago, writing shitty blogs for content mills, as we all do. And then I slowly moved up market. So that's a little bit of a background and it's not unique to me at all.

Dane: We don't have enough time for all that. Yeah. Well let's get into that. I'm really curious how you're seeing the evolution of content, like now that we realize slop is a thing and that slop is bad. There's been this wave of people hiring content writers to really do the human end of things. How are things going with you and how do you see that sort of evolution right now? Or is that, have we decided slop is bad, or are there still people that are still in the slop camp?

Brinda: So slop is bad, yes, but saving money is better, right? That's always the case. But I am observing a fork in the road. So I have some clients who are going all in, automating the entire content life cycle. And after six months, they come back and they're like, actually, we would like a human in the loop. And the other clients have been very steadfast on maintaining premium quality content with a strong narrative voice, original research, all of that from the get-go. And they haven't experienced those fluctuations in rankings or panicking after every Google update, or every time Claude rolls out a new model, they've decided that this is the time we fire our marketing team. So I'm observing two kinds of brands here. Some of them are trying to do a middle way, but I feel the more they get into AI, they feel that the first output is good, the second output is good. The problem is with repetition. So across two outputs, you can get a good enough sort of thing for yourself. Across a hundred, that quality doesn't hold, it breaks.

Dane: Right. This is an interesting point about the middle road. I want to get into that a little bit. One of the things I've been really interested in and experimenting with is using video interviews as the fodder for written content. So we're doing this talk right now, so there's gonna be a transcript that's available. What's the best way for us to take that transcript and then give that to a writer to write quality, like human content, with that as an ingredient, the transcript, but then we're also getting video content out of it. I think writers and people like me, video content producers, are natural allies. If we can find the right way to do this, because written content on its own isn't quite as powerful as written content that also has a video in it, right? It's gonna rank better in AI search. It gives people another way to consume information. Some people want the videos, some people want the writing. It's kind of all over the map about how that goes together. So do you have any thoughts or experience with the right way to use transcripts? If someone gives you transcripts, what's the right way to do that, to squeeze the most juice out of that? As a video producer, if I'm working with a writer, how do I best set us up for success by giving you the thing you need to make great written content?

Brinda: I am working and I have worked with a lot of transcripts as raw material purely and then pulling out the insights and turning that into written content. My first port of call, and I'm sure a lot of people will disagree because I do not feed it into AI first thing. I do not ask it to summarize it for me. I've noticed that if I make the laborious choice of going through the transcript by myself, page by page, I can catch certain phrases and turns of phrases that are very unique to the speaker. And then perhaps I can go into Claude and say, I found something really interesting that this person said, a terminology that's unique to them. Can we perhaps, or are there any other arguments in the entire interview that support this hypothesis or no? So I use AI as a rubber duck, which is basically you keep asking it questions, it asks you questions back, and in that process, you find answers for yourself.

Dane: So you're kind of hinting at something that I have always felt, that because the way that people speak is so natural and authentic, that is the real value of the human part of all this. AI will summarize stuff until it's all middle of the road and it takes out all the personality. But if you can, like you said, get those little sound nuggets, sound bites, word choices, whatever they use, that's the stuff that I think is gonna really help people stand out. So there's some secret ingredient or secret process here where you're taking those transcripts and picking out those beautiful gems, then doubling down on that and blowing those up to be deeper, more valuable. So maybe the perfect scenario would be I do a video interview with an expert or ten or a hundred experts, whatever it is.

Brinda: Yes, indeed.

Dane: And then I give you those transcripts and you're going through picking out those gems. And then somehow we're using AI to combine all those into one bank, one beautiful box of jewels, and then we compress those together into the most high value stuff. And maybe then make another video about it that takes all the good stuff and puts it in one thing.

Brinda: Cohesive argument, yes. I mean you can make endless content if you know what to look for. This is why I don't like to just be called a writer, because when I'm going through transcripts, it's a repository of judgment. So it's up to me what to exclude and what to include. And that discernment, that taste comes from years of practice and looking at what really bad content looks like and sounds like. To then compare it to what really good content could sound like. So when you're giving me complete raw material, I go in research and strategist mode. The writing comes as the last step, which may be counterintuitive to a lot of people's processes. But to me, I feel the writing is perhaps the easiest step after I've compiled everything I need to know. I studied creative writing for five years, and my professors would always tell me that you don't sound smart because you use big words. You need to have a coherent argument, something that can carry from the first sentence to the very last, and you would still be prompted to read. So that kind of judgment that pieces everything together is not only writing. It's strategy.

Dane: Yeah, this is a great point. I've been really focused on content strategy for a while now. For my 30 year career in video, most of that was being somewhat of an order taker. Can you make me a video, and then I go make the video? But there's already been a lot of decisions that have been made, strategic decisions about what to do, what to not do, what to say, what to not say. So I'm really advocating for getting the strategy part earlier in the video process. Whenever someone says the word video, I always want to say why. What are we doing in the first place? And then, like you said, the real value is the thinking, the judgment, the taste that happens earlier in that process before you get to writing or recording video, which are sort of these end-of-stage steps where you're taking all these ideas and decisions and the writing or the scripting is just putting those ideas together into a coherent package.

Brinda: Yes, I've been thinking more about this as well. And I feel like if you give a senior writer or strategist the same transcript and you give someone who is new to content marketing the same one, the outputs, even if you tell both of them to use AI, will be wildly different. Simply because sometimes a strategist also knows when to use jargon to connect to the audience. That's also a decision. But if you go by blanket rules and just based on a prompt like remove all jargon, pick out gems, give me the most impactful summary, that's not gonna produce anything of value that hasn't been said a hundred times over all over the world.

Dane: Yeah. I think we're at this point now with AI search where it's becoming more and more obvious, at least to me, that the thing we need to be producing are the direct answers to the questions that people are asking. And like you said, pulling out those gems that are judgment, taste, judgment calls about when someone's trying to make a buying decision, this or that. We need to have a system to get those gems out there consistently because as the market keeps changing, we're gonna have new insights that need to be collected and surfaced to help these buyers make buying decisions along the way. So how can writers and video content producers help make a system to get these subject matter experts to drop these gems out there in the marketplace more frequently? Do you have a process for this?

Brinda: Okay, I'm gonna plug my framework here. This is a very natural plug for my framework. It's called ARS. Some people like to call it ARS, I don't mind. So it stands for accessible, relatable, and specific. If you can make the language accessible, if you can make the emotion relatable and the example specific, one of each, just to begin with, for a keyword, let's say, and then move from there, you have much more clarity about what you're gonna say to your buyer.

Dane: This is good stuff. I like it. It's another way to think about it. You're kind of dogfooding your own approach here. We need to have a way of getting these gems out of people, and so your framework is a good one. And it's unique to you, right? So now you're gonna be kind of known for this, so you're illustrating exactly how original thought leadership can be surfaced as people are trying to solve these problems of how do we show up in the first place. Well, by having an original thought that has some judgment behind it.

Brinda: Yes. So even for my SME interviews, which I do for a lot of my research, I follow my own framework. So of course, you ease them in and everything. I use very plain language, but I also use jargon that's very familiar to them. So I will say streamline processes and everything because that is familiar to them. That's part of being accessible to your audience. So that's also a judgment call. When we're talking about relatability, I write about stuff that I've been through myself. So I've run an Instagram thrift store, for example. So I'm now very deep in the world of e-commerce. So I'll say something like, okay, when I was running my thrift store, Instagram marketing was behaving this way. What have you seen? And then for specificity, I ask them for one example which they remember from their career or the last five years that's had a measurable impact on their business or their personal life that's worth talking about. This is obviously a very broad overview, but I feel with the framework you can ask more targeted questions. More targeted questions obviously lead to better quality answers. And the answers, which become the raw material of that transcript, give the writer or the strategist something concrete to work with. Because the main thing is pulling out original thought.

Dane: Right. And I'm getting curious about your thoughts on the frequency of doing this because things are always changing. We need to have fresh content. It can't just be our original thoughts from two years ago. So how are you looking at your recommendations for how frequently you're posting new content, harvesting it through interviews? What's a good cadence to stay relevant and top of mind and visible?

Brinda: I know it's not budget friendly, but I would say six months.

Dane: So what does that look like? I'm a B2B company and we're in growth mode and we want to get our message out there and answer buyer questions. So what's the process look like for harvesting content for you?

Brinda: I mean it all depends on the founder's vision, the content strategy that's already in place, or whether I have to make one from scratch, the distribution channels we're targeting, the audience, of course, where they live, and if we can extract what their pain points are. So that's a much more long-winded process. But I would say that I have seen companies simply update the date to signal freshness. And it backfires because I can tell that you're using old statistics, you're using old examples, you're using government laws that are not in effect anymore. So I would say six months for the very reason that we all know AI is moving at breakneck speed. Companies, meanwhile, are also trying to keep pace and the features are just multiplying. Like every software you subscribe to now, we have a ChatGPT connector, now we have a Claude connector, it's impossible to keep up. So I wouldn't say that the goal here is to be perfect with the updates, like you detail everything in a change log that's separate. Your content doesn't need to reflect that, but it does need to reflect the most relevant recent changes that are relevant to your buyers.

Dane: Yeah, right. Okay. Well, this has been a great conversation. As we wrap up here, do you have any sort of final advice or a tip that you want to give out to potential people that are struggling with their content?

Brinda: Yes, read fiction.

Dane: Good advice.

Brinda: Yes, absolutely. A lot of us tend to optimize everything, even our reading habits. And that sort of quells original thought in itself. Because if you're working towards a measurable goal, you have something you wanna hit, it becomes another thing that you're tracking, optimizing, analytics. Like I read this many books per year, I read this many marketing books per year. So I would say it's good for your brain to have different tracks when it comes to a discipline. You can read marketing books, sure. You can read brand books. Try reading science fiction in the meanwhile. It gives you variety in language, it gives you variety in thought, and it also shows you different ways of thinking about the same thing.

Dane: That's great advice. I love it. Well, Brinda, thanks a lot for sharing your insights today and wish you best of luck out there.

Brinda: Thank you, Dane.

Dane: We did it.

Work with Digital Accomplice

I help B2B companies grow visibility, trust, and pipeline with strategy-led video content — strategy, creative, production, and distribution, all in one place.

Book a meeting → Subscribe to my newsletter for B2B video insights →

Dane Frederiksen CEO / Creative Producer www.digitalaccomplice.com · dane@digitalaccomplice.com