Speaking with experts who love their jobs usually makes you aware of two things simultaneously: 1) how much you don’t know, and 2) how much you’d like to know. My conversations with Prashant Darisi and Vick Vaishnavi were no exception. I walked away with an appreciation for the decades of expertise, knowledge, and experience behind Octave’s key players, and a more solid understanding of how these elements have developed a human-centered AI platform that doesn’t replace human potential so much as magnify it.

This, in a world where many big companies are tempted to prioritize monetary gain as they skim over the details, is not only a point of principle but a distinct advantage in terms of quality.

How? For one thing, there’s no shortage of people providing the data and context to keep AI in check (ask Vaishnavi), and for another, the tasks involved are inherently more connected and supported by a wide range of professionals and operational use standards than ever (ask Darisi).

I began with general questions. 

“After listening to the keynote talk, I have to say there was a lot of discussion about the responsibility of the tech industry, manufacturers, and AI, and how they’re going to interact now that we’ve got the ball rolling. One thing I thought the keynote addressed well was using AI as a tool—not to replace human expertise, but to augment it. I guess my biggest question for you is: How does Octave use AI to elevate human capacity rather than replace it?”

It’s not meant to be adversarial, but guiding. Darisi answers by highlighting the importance of setting the stage.

“I think for us, it starts where every software company has been for decades. What person and what customer are you serving, and what outcome is the customer expecting?”

For an idea to be beneficial, it must complete the expected tasks with competence, Darisi says. “My theory is very simple. I’m driving a Toyota Corolla. Maybe I’m not very proud of it and I need a Jaguar or a Porsche. If my GPS is wrong, a Porsche will only take me to the wrong place faster. That’s fundamental.”

‘If my GPS is wrong, a Porsche will only take me to the wrong place faster.’—Prashant Darisi

On a similar note, Vaishnavi emphasized that context is vital to both accuracy and completion of tasks.

“I’ll give you an example,” he says. “A few years back, not too long ago, everybody used to talk about content. Content is key, right? With AI, it’s the context that’s key. It’s about context. So if you take the human out of the loop, you’re going to lose the context.”

He mentions the keynote address.

“What you also heard Mattias (Stenberg, Octave CEO and the keynote speaker) talk about was AI with consequences, right? And AI does have consequences. So, it’s like doing something faster, more efficiently. But if what you’re doing is not desirable, then you just compounded the problem by doing it faster and at a bigger scale.”

In essence: garbage in, garbage out. All AI does with garbage content is magnify it.

“We believe, and our customers do too, that every time the use of AI is described you’ve got to have a human in the loop. It’s not about distrust of AI, but the context. The context checking is always human intervention. Like, hey, am I doing it in the right context? Am I just doing it blindly?”

Vaishnavi speaks candidly on how and why AI is useful.

“It’s a productivity exercise,” he says. “It’s an ability to scale without having too much human effort. But it doesn’t, certainly in our philosophy and the way we look at it, replace the human. It just augments human efforts and makes them more productive in a positive way.”

Darisi also notes that AI at Octave is more than just another fancy tool. After all, what good is a tool if it doesn’t add intrinsic value?

“This assumption,” Darisi says, “that vendors will build more bells and whistles only to take the customer to the wrong place faster has been true for decades. That philosophy, for me as a product leader, hasn’t changed. So, we look at the customer and ask, ‘Who are we serving, and what are we trying to accomplish?’ And if you look at a manufacturing or industrial setup, you have customers operating at the shop floor and customers operating at the top floor. What value and business outcomes are they looking for?”

Octave and productivity

Whether it’s auditors, HR personnel, or people on the shop floor, everyone has the same goal: productivity.

As Darisi puts it, “They’re looking to avoid mistakes. They want to make sure the work they’re doing is efficient. If things are changing, they’re looking for knowledge. They’re looking for checklists. They’re looking for information.”

But the catch is that there’s a constant influx of changing information, training, and standards.

What workers are asking for, Darisi says, is as simple as, “How do you make my job better? Either make me faster or help me do it better so I don’t get pushback. If things have changed, how do I have that knowledge at my fingertips?”

This he calls the connected worker—and it’s what Octave is trying to support and produce more of.

“One of the biggest problems this industry is facing is the shortage of skilled workers at the shop-floor level,” Darisi says. “So we’re delivering knowledge to them. A worker can take a picture or video and ask, ‘Did I do this right?’ AI can use OCR and image recognition to say, ‘No, that’s not how the part goes,’ then immediately deliver a two-minute training video.”

Another Octave capability is Form Field Advisor.

Darisi says, “People are constantly filling out supplier corrective actions, preventive actions, nonconformances, audit findings, document control forms. How do you know you filled it out correctly? How do you know your priority is right? How do you know you selected the correct workflow or the right department?

“A lot of back-and-forth happens simply because people fill out forms incorrectly.”

Before long, complaints are flowing in faster than solutions. This is a common problem for people in all industries—and one that AI can begin to address.

Darisi then walks AI “up the food chain.”

“You’re measured on outcomes like defect rates, scrap reduction, recalls, leading and lagging indicators. But ultimately, your customers have to buy your product.

“Somewhere, you need to connect customer experience to the work quality professionals are doing.”

The challenge is that customer complaints don’t happen in one email. They span multiple tickets, photos, and receipts, and it takes time to put everything together to form a coherent picture.

“Your response to the customer changes completely,” Darisi says. “You’re no longer investigating. You’re telling them (based on information you’ve now uncovered) ‘Return it to your distributor. They’ll replace it with the corrected inventory.’”

Darisi also points to audit risk.

“One customer has 40,000 documents loaded into the system. A single standard may affect hundreds of documents, because requirements overlap across HR, safety, best practices, and regional regulations.

“AI can identify conflicting language or missing requirements before an auditor does.

“To me, AI has to walk up the food chain. It has to help shop-floor operators, predict customer behavior, surface audit risks, and make better use of organizational knowledge.”

Content, context, and consequence

Vaishnavi says, “The positivity is described by the context, which only the human has. Keep in mind that AI is very highly driven by data that it consumes. If you give it garbage, it’s going to produce garbage. It’s just going to do more garbage faster, better, at scale.”

“Yeah, sort of magnifying the quality of what’s already there,” I respond. “That was the big takeaway for me from Matthias (the keynote speaker), from the conversation with Prashant, and just that we are supplying the data, which is a large part of that context.”

“The domain knowledge, the domain intelligence that you heard,” Vaishnavi says, “that only comes from the humans. They understand the context. I always say, it used to be about content. Content is kind of synonymous with data. You need context, which is the human. And then you apply AI to it, and you get the consequences. So it’s content, context, consequence. That’s why you don’t take the human out of it.”

I ask how Octave connects people, processes, and data so that it seamlessly transforms workflows.

Vaishnavi says, “If you think about any enterprise as fundamentally people, they have processes that use the people to do something, and they have technology that they use to implement or digitize the process and make people more efficient.

“What Octave essentially is doing is bringing these three things together in a closed-loop model, ranging from when you design any type of asset. It could be a facility, it could be a building, it could be a product, it could be a widget, anything.

“Second portion is, how do you build it? Meaning, you’re going to either construct it, or you’re going to manufacture it, one of the two, or you’re going to assemble it.

“How do you operate the finished product? Again, that could be a building, it could be an asset, it could be a widget, it could be a manufactured goods, and then protect it.

“The trick, essentially, is not just doing each of these functions independent of each other. They’re not mutually exclusive. They’re actually intertwined.”

In manufacturing, something produced as part of the process might develop a defect.

Vaishnavi says, “You may do a root cause and realize, hey, it’s how we built it and assembled it that was a problem, the way we were doing it to specs. And then the build guys will say, geez, we only did what you told us. It was functions as design. So you know what? It’s really a design problem. So to navigate this little loop, if you will, between the design, build, operate, protect, you really want it to be a continuous improvement cycle.”

At Octave, Vaishnavi says, “We really want to supply the intelligence, in this case, again, the data, the context, and the consequence.

“The decision that you made here can have a ripple effect downstream and therefore has an impact on the quality of the defects that you produced. So, you need to be able to look at the entire process holistically in every stage and do it in real time, so to speak.”

As something goes wrong in the operate mode, the information can be sent back to the people responsible for build and design.

Vaishnavi says, “So the idea is to have a turnaround cycle, which is faster, as opposed to, well, ‘We only learned about the problem after it went to the customer.’

“Because the cost of repairing it or remediating it at that point is, you saw the numbers, that 1x, 5x, whatever, 25x, 100x, is to collapse that cost of correction, if you will, or remediation, do it more upstream before it actually leaves your manufacturing facility.”

Hindsight, insight, and foresight

I respond, “I think that hits at an interesting point, because people talk about predictive maintenance, but there’s also that piece of figuring out where something is in the process of being made, figuring out the flaws in the process. And that’s before you’ve rolled out an entire product or gotten customer complaints.”

“I like to think of, and the use of AI is important here, basically three types of optics you are getting,” Vaishnavi says. “Historically, everything used to be what I would call hindsight.”

“Yeah, of course, hindsight’s always 20/20,” I say.

“What you just described is insight,” he says, “which is in process, tell me what’s wrong in real time so I can fix it.”

“The third one is foresight, which is to look at how we have functioned over a period of time. Again, collect your content and data over a long period of time, analyze it, and decipher and uncover what could go wrong, as opposed to before it goes wrong. So again, I would say hindsight, insight, and foresight. Historically, everybody has had hindsight, and it’s easy to claim that. It’s the insight, and then more importantly, the foresight.”

Building a bridge of intelligence

I ask what problem Octave is trying to solve that is fundamentally bigger than quality management.

Vaishnavi says, “If you look at the four phases in running a business, the design phase, the build phase, the operating, the production, each of these, think of them as islands where coordination is being applied.”

“Think of these as four distinct islands of automation. The difference is, what Octave is doing is, it’s building a bridge across these four islands, so that while each of the islands can automate into themselves and become really good at designing, really good at building, the fundamental difference is, it’s building that bridge of intelligence that allows each of the islands to leverage automation, but make sure it gains the insight and the foresight to make sure that it doesn’t propagate to the next phase, something that causes a defect or an impact, let’s call it negative impact, to the business before it happens.”

I ask, “So how do you bridge the gap, no pun intended, between these four islands?”

Vaishnavi says that it all comes down to decision intelligence, or understanding how to minimize the effects of unknown variables by providing information (aka context).

Decision intelligence

“Again, this goes back to AI and the human,” Vaishnavi says. “Who makes a decision? Do you really think about it? Like, would you really want to trust AI to make all decisions for you? Probably not.”

I discuss times when I use AI vs. times when I rely on my own knowledge or that of colleagues to make a final decision—writing vs. catching typos, for example. I have a voice, a cadence, and expertise for how my content, style, and sources ought to interact in a way that optimizes the experience for a reader and remains my work. The decisions that require knowledge of nuance and ambiguity require more humanity behind the wheel, in a sense.

Vaishnavi agrees. “I always say, they will not ever outsource their decision making or decision intelligence completely to an agentic AI system. It’s there to help me make the decision. It’s not to make the decision for me.”

I discuss the value of experience in making decisions, citing a quote often attributed to Mark Twain: “Good decisions come from experience. Experience comes from making bad decisions.” Vaishnavi agrees that there’s a piece where people need to be able to look at the failures of the past and the data that those failures have given them.

“Feed that to AI, and say, hey, tell me what lesson have I learned. I always say the most powerful element in this universe is time,” Vaishnavi says. “And it only gives you two things: lessons and experience.”

“And so, even an agentic AI has to be fed that, to your point, the lessons and the experiences, which is data, to say, hey, can you tell me what I should do? What it is doing is, you only have one brain and it can process at certain cycles per second. It’s just doing that at hundred or thousand times the speed and scale for you. So it’s a tool for you.”

He sums up the value by incorporating time and analysis to the mix.

“So what would have taken you weeks to kind of analyze that big amount of data, it’s just doing it quicker for you. But make no mistake, you’re not going to outsource your thinking to that agentic AI system.”

But it gives that expert more time to take on other issues.

“Other issues, and maybe tackle more problems, first of all,” Vaishnavi says.

“If I didn’t have it, and I was only focused on one, I could only do one thing at a time. Now I can do three things. So I get concurrency of my time. I get my time back, which is more valuable to me than just having an agentic AI.”

‘So I get concurrency of my time. I get my time back, which is more valuable to me than just having an agentic AI.’—Vick Vaishnavi

Decision automation and decision acceleration

I ask what types of decisions should never be fully given over to AI.

Vaishnavi says, “The way to create a framework for that is to have a mapping between the area that you want to make the decision on, or the topic, proportionate to the risk it presents to yourself.

‘I’m never going to outsource my decision.’—Vick Vaishnavi

“There are many people who can quantify risk on a scale of zero to 10. Like, hey, above eight, I’m making this as an example. Above eight, I’m never doing that. If it is below two, yeah, let’s make the decision. How bad can it get? We can absorb the risk.

“The time I would expend to make a decision on a low-risk topic is more valuable to me than the risk itself, that the damage it would cause. It would not have a material impact on our business, but if it is eight and above, I want to be involved. I’m not letting that system take that decision.”

The other consideration is what Vaishnavi calls the repetition index.

“If there’s something that happens day in and day out, and it is within certain guardrails of control limits, it always works within this limit, low, high, low, high, low, high, and you have enough data available for weeks, months, years, whatever.

“If the predictability index of that decision is always gonna be above eight, it’s always gonna be the same thing, yeah, let that thing make a decision. As long as it’s between those two limits. So one is risk-based, and the other is repetition index, or predictability index.”

The idea is that when risk is low and predictability is high, AI can make the decision. When risk is high or predictability is low, Vaishnavi still wants human involvement.

“I might use it, but the final decision is still mine.”

Later, he describes these roles as decision automation and decision acceleration.

“Decision automation, in cases where the risk is low, predictability is high, OK, let it be automated. Where the risk is high and the predictability is low, I want decision acceleration, but this guy can help me accelerate my decisions.”

Quality as a strategic advantage

Manufacturing already uses different systems to deliver a final product.

Vaishnavi says, “You have PLM systems, that is really for product life cycle management. You have an MES system that’s, oh, how am I gonna actually do it? You know, PLM told me, this is what you’re gonna build. MES tells me, how do I do it?

“Then I have a CMMS, perhaps, that’s like into the production piece. And then I have a QMS that tells me, oh, your quality’s up, down, etc.

“These are all different islands of automation that are available. But there hasn’t been a bridge built between all these islands, which would allow manufacturing industries to essentially look at quality from an end-to-end perspective.”

That bridge is supported by the amount of domain-specific data Octave has accrued over decades.

“So, the context,” I say.

“The context, exactly,” Vaishnavi says. “We have the ability to provide the best context there, since we are sitting on the data. What you heard yesterday is, how are we going to kind of turn that data into intelligence so that decisions can be made? Your decision, as I said, is only as good as the data you have. So the richness and the longevity of our data gives us a huge advantage.”

He notes that the context at Octave is valuable precisely because of that history.

“That was accrued over decades of time. Our competitors cannot fast forward time and accrue that. The time works at its own pace.”

Ultimately, those decisions reflect in quality.

Vaishnavi says, “If you don’t have a way to measure, manage, control, and enforce quality, by design, again, you’re doing the middle pieces, but you’re not doing the fundamentals of that mixture. We look at quality as a strategic advantage. Quality will remain the centerpiece because it’s the byproduct of everything you do.

“You have to look at quality as a strategic advantage or a business differentiator, as opposed to tactical necessity. If you want to be differentiated in the business, quality is your best friend.”

‘You have to look at quality as a strategic advantage or a business differentiator, as opposed to tactical necessity. If you want to be differentiated in the business, quality is your best friend.’—Vick Vaishnavi

And if time is the most precious commodity, one of AI’s central benefits is reducing the time quality professionals spend analyzing different kinds of data.

Vaishnavi says, “Because there are so many functional elements of quality, whether it’s an audit, or it’s compliance, or it’s in training, etc., it’s the analysis that they have to do on the data set that they accrue.”

AI can help with that analysis, connect the information, and accelerate a decision. The content comes from the data. The context comes from people. And then you apply AI to it, and you get the consequences.