Streamline & Scale Newsletter: Issue 8: Podcast

Welcome back to the Deep Dive. Today we're doing something a little different.

Yeah. A bit of a reality check. I think

usually we look at where technology is going, the next big thing, right?

Yeah.

But today we need to stop and look at where it's sort of stalled and how it's finally starting to move again.

Absolutely.

We are looking at the latest issue of Red Hawk Technologies newsletter Streamline and Scale. This is volume 2 Issue 8.

So when I saw the topic of this issue, it felt, well, almost retro.

You're talking about the title AI Chatbots when they deliver ROI and when they don't.

Exactly. I mean, we've been having the chat bot debates since what, 2018?

And then again with the big generative AI boom in 23.

Right. It feels like we should have solved this by now. But Red Hawk is arguing that for a lot of businesses, the ROI. It just isn't there. Or at least it's not there automatically.

That's the key distinction. I think. We aren't questioning the, you know, the [00:01:00] capability of the tech anymore.

No, not at all.

In 2026, the models are brilliant. The issue, and this is really the mission of our deep dive today, is figuring out why we still have companies deploying bots that make you just wanna scream representative.

At your phone.

It's that disconnect. Yeah. The promise versus the little pop-up window.

Precisely. So to get to the bottom of this, we've got a really interesting stack of sources. First, there's the editorial from Red Hawk, CEO, Matt Strippelhoff,

which really sets the stage on the strategy side.

It does. Then we have this very blunt article.

It's called Our Chatbots. Still Worth It. Here's the Honest Answer.

I do love a blunt article title.

It doesn't pull any punches. And then my favorite part. An anonymous case study.

Ah, the anonymous ones are always the best.

They are

the official ones are always so polished, you know, everything went perfectly.

The anonymous ones, they tell you where the bodies are buried.

And this one is a proof of concept story. It shows how a company went from like total skepticism to actually using this stuff. [00:02:00]

So let's start with that big elephant in the room. The newsletter tackles. This chatbot is dead idea right away. Is that really a thing people are saying now in 2026?

It is, but it's cyclical. The source material points this out beautifully.

Okay.

Every couple of years the tech world declares chatbots dead first. It was because those old decision tree bots were just too rigid.

Oh yeah. The ones were, if you misspelled one word, the whole thing would just break.

The whole conversation just restarts.

Exactly.

Yeah.

Then it was because the early generative ai, you know, the hallucinations were just too risky.

Right. The lying bot problem. Yes. We sell flying unicorns. Would you like one

Exactly. But the argument here is that the chatbot isn't dead. It's just graduated. We've moved from those clunky scripts to genuine AI assistance, but, and this is the key.

Business leaders are tired.

They've been sold the dream three times now,

this time it really works, we promise. And that fatigue, it [00:03:00] leads to skepticism. So the Honest Answer article gives this, well, it's like a litmus test. It says, stop asking if chat bots are good. That's the wrong question. Ask if your business actually needs one.

It had that checklist, right? For, for ROI.

It did. Yeah. Three things. Volume, data, and purpose.

Okay, let's break those down. Volume seems pretty straightforward.

Volume is the easy one. If you're answering, what are your hours? 500 times a day, sure. Get a bot. It's just simple math,

right? Save your human hours.

For human work.

Exactly. But the other two, purpose and data, that's where people trip up. Let's talk about purpose.

Okay.

The article argues that too many companies deploy ai just to check the box,

the innovation theater approach. Yeah,

it looks great. On a slide deck. Look, we're innovative, but without a clear purpose.

Reduce ticket resolution time by 20%. It's just an expensive toy.

And that leads to the biggest trap, the one they mentioned in the sources, the set it and forget it mentality.

Oh, this is where that Commonwealth Bank story [00:04:00] comes in.

Mm-hmm.

I've read that and just cringed.

It was painful. It really felt like watching a car crash in slow motion.

It's such a good cautionary tale.

So for anyone listening who hasn't read it, Commonwealth Bank rolled out a new voice spot. The goal was pure efficiency. Cut costs

aggressively. Yeah, they actually announced job cuts before the bot was even proven. They were banking on it, taking over the workload,

and what happened?

The total disaster. The bot just wasn't ready for the nuance of real banking questions.

So people got frustrated,

frustrated, hung up, and then immediately called back trying to get a human. So call volumes didn't go down.

They spiked.

They spiked. They created so much more work.

Wow.

They had to reverse the job cuts and actually hire more people to clean up the mess.

Which brings us to what the newsletter calls the empathy rule. I thought this was a really useful way to frame it.

It's a great rule of thumb. It's basically a sliding scale.

Okay.

If the customer's interaction needs empathy more than it needs efficiency, you just. You don't [00:05:00] use a bot.

So gimme an example of the empathy side,

uh, a healthcare diagnosis.

Yeah.

A complex legal issue or just frankly, if a customer's already angry. When someone's emotional. A robotic, I do not understand is like throwing gasoline on a fire. It feels so dismissive

for sure. But on the flip side, if I just wanna know. Where's my order?

You don't want empathy. You don't want a conversation about how my day is going.

I just want the tracking number.

You want speed. That's where the bot shines. So it's about knowing if your customer's in transaction mode or crisis mode.

That makes perfect sense. Okay, so that's volume and purpose. Let's get to the middle one data, because the newsletter argues this is the main reason why even smart bots feel dumb.

This is the data disconnect, and this is, I think, the real technical heart of the whole problem.

Explain that because in my head, a top tier AI model in 2026 has read the whole internet. It should know everything. Right?

Okay. Think of the AI model like a, a really smart, [00:06:00] charismatic, new employee. You know?

Right.

Let's call him. I don't know. Chip.

Chip, okay. I like chip.

Chip has a PhD. Communicates perfectly. He's read every book, ever written. Great hire.

Sounds amazing,

but it's chip's first day at your company and you haven't given him a login. You haven't shown him your product manual. You haven't introduced him to a single customer.

So Chip can talk about Shakespeare, but he doesn't know anything about us.

Exactly. A customer asks, where's my order? Chip can only say, ah, shipping is a fascinating logistical process.

Yeah.

He can't actually look it up.

That is the dumb bot experience. It gives you the generic policy instead of your answer.

Right. A good bot, according to the source, is integrated. It's plugged into the CRM into the inventory system. It recognizes you. Hi. Are you asking about that red sweater you bought last week?

That seems so obvious. Why is that still so rare?

Because integration is the hard work. It's the unsexy part.

The plumbing.

It's the plumbing. It's much easier to design a cool chat bubble [00:07:00] the interface. Yeah. Than it is to wire the bot into a 20-year-old database, the foundation.

So companies are buying the Ferrari body and putting it on a lawnmower engine,

and then wondering why they aren't winning the race.

This is a perfect segue into that case study because this manufacturing company, they had this exact problem, but the stakes were so much higher.

Way higher.

Hmm.

This wasn't about selling sweaters. This is a North American specialty manufacturer In a highly regulated market,

regulated is the key word there.

It's everything. If their bot gives a wrong answer about, say, a chemical composition or a safety protocol. They don't just lose a customer. They could get sued.

They could face massive fines.

So hallucination, the ai, just making stuff up was a complete non-starter,

zero tolerance. They needed a precise ai, not a creative one.

And what was the specific pain point they were trying to solve?

That's what the source calls the access problem. It wasn't that the company didn't know the answers, the knowledge [00:08:00] existed, it was just trapped.

I think every single person listening can relate to that.

Oh, absolutely. The answers were buried in, you know, PDFs on a shared drive, old email chains, and the worst one. Institutional memory.

Ah, yes. The Bob Factor.

The Bob Factor. Let's say Bob has been there 20 years. He knows everything. Yeah.

Bob has to eat lunch eventually.

Or go on vacation,

or heaven forbid, Bob retires

and all that knowledge just walks out the door.

Exactly. So their customer service reps, their CSRs, they were drowning. They spent all day just hunting for information while a customer was on hold.

It's slow, and I imagine the answers were inconsistent,

totally inconsistent.

So leadership says we need ai, but they were terrified of the risk. So they brought in Red Hawk for a proof of concept, a PoC,

and I love how the newsletter frames this. It says, this wasn't a science experiment, it was a risk decision. What does that mean?

It means they weren't just playing with New Tech for fun.

They were testing a business hypothesis. Can we build a system that [00:09:00] is safe enough for our standards?

So they decided to build it for internal use first.

Yes. A tool for the CSRs, not for the public

smart. It's like a co-pilot for the support team Let's the humans verify the AI before a customer ever sees it.

It reduces the risk profile immediately.

Okay, so let's get a little technical. How do they actually build it? The newsletter mentions. Chat with your data and this acronym, RAG,

Retrieval-Augmented Generation. Sounds a little clunky.

A little bit, yeah.

But it's actually the architecture that makes all of this enterprise AI stuff possible and safe.

Okay. Break it down for me. How is it different from just, you know, asking a normal AI a question

with a normal ai? The model answers from its general training data, basically the internet. It's broad, but it can be shallow and wrong. With rag, you add a step in the middle.

The retrieval part,

exactly. Think of it like a court case.

The AI is your lawyer, but you don't let the lawyer just make things up. You hand them a box of evidence, your company's [00:10:00] PDFs, your safety sheets, your emails, and you say, answer the question using only what is in this box.

That's the grounding. You're giving it guardrails.

Perfect word for it. So in this case study, Red Hawk used Microsoft Azure.

They ingested a year's worth of messy internal data

and indexed it all.

Yes. Using Azure cognitive search, and this is important, it uses semantic search.

What's the,

it understands meaning not just keywords. So if you search for broken screen, it knows that's related to crack display, even though the words are different.

It finds the concept,

okay, so the system finds the right documents

and then it sends only those specific chunks of text to the AI model and says, read this and answer the question,

and this is the critical part. It cites its sources,

the Zero trust principle, the bot gives you the answer and it gives you a link right back to the original document.

So the CSR can say, okay, the bot says X. Let me just click here and verify that

it changes the game from trust, the magic box to trust but verify,

which is exactly what a regulated industry needs. It turns the AI from [00:11:00] a creative writer into a diligent librarian.

So the big question. Did it work?

It absolutely did.

They stress tested it on a really complex product line. The bot was answering safety and troubleshooting questions instantly.

Wow.

It was even pulling up historical resolutions. You know, we saw this exact problem two years ago and here's how he fixed it.

So Bob can finally eat his lunch in peace.

Bob can eat his lunch.

And even more important, the new hires have access to Bob's Wisdom on day one,

but beyond just the speed the newsletter says. The big win was confidence.

That was the huge takeaway for leadership. They sat down, they tried to break it. They asked a tricky questions, and because of that rag architecture, it didn't hallucinate it either gave the right answer or it just said, I don't know.

I don't know. Which is a perfectly acceptable answer in business making Stuff up is not,

it's a safe failure mode. There's a quote here from Matt Strippelhoff, the CEO, that I think just sums it all up. He [00:12:00] said, chatbots were never the problem. Strategy was,

that's a strong statement.

It really is. It implies we've been blaming the tool for the carpenter's mistakes this whole time.

I think we have, we've been buying the tool and just hoping it builds the house by itself. Red Hawk's approach is all about something they call operating leverage.

Unpack that for me. It sounds a little like finance jargon.

It is, but the concept is simple. It means decoupling your revenue from your headcount.

Usually in a service business, if you wanna serve twice as many customers, you have to hire twice as many support staff. That's linear. It's expensive. Hmm.

Operating leverage means I can double my customers without doubling my staff.

Exactly. Because the AI is handling the friction, the searching, the summarizing the easy questions.

Your humans are still there, but they're doing human work. They're handling the empathy cases, solving the complex problems.

They aren't just human search engines anymore.

You are taking the robot out of the human. You aren't replacing the human, you're upgrading their job description. [00:13:00]

Looking at the big picture, then it's 2026, the tech is ready.

This architecture, it works. The risk can be managed. So why isn't everybody doing this?

I think it's fear.

Yeah,

and inertia and honestly just the messiness of their own data.

It always comes back to the data.

It always does. Red Hawk actually includes a call to action here because they know this is where people get stuck.

They offer a free AI discovery session and a PDF assessment.

Basically a way to figure out, is my house even in order for this?

A much better first step than just signing a million dollar software contract. You need to know if you have the ingredients before you try to bake the cake.

Before we wrap up, I have to ask if I'm a business leader listening right now and I'm still on the fence, you know, waiting for the next version of ai, am I making a mistake?

I think you are. That wait and see strategy made sense in 2023, maybe even 2024, but in 2026. The wait is over. The bottleneck isn't the technology anymore.

So what is the bottleneck then? [00:14:00]

The bottleneck is your filing system. I mean, honestly.

Seriously?

Yeah. The bottleneck is the fact that your company's most valuable intelligence is locked in someone's head or in an email from 2019.

So while you're waiting for the AI to get smarter, you should be getting your data cleaner.

That's it. Exactly. The smartest AI in the world is useless if it can't read your company's book. And right now, for most companies, that book is closed and collecting dust. The tech is ready to read. You just need to open the pages.

That is a fantastic point to end on. The robot is ready. It's just waiting for you to hand it the manual.

Yeah, couldn't have said it better myself.

Well, a huge thank you to Red Hawk Technologies for this issue of Streamline and Scale. And thank you for tuning in. Go check your data and we will catch you on the next deep dive.