Cameo Doran

Chief Product Officer ⇢ Fractional Prices with Full-Time Impact ◆ Digital Transformations ◆ Advising Startup Founders and Executives Building Innovative Products ⇢ AI / XR / Web3

I’ve always been fascinated (and a little frustrated) by how many great companies can’t seem to ship great software. I’ve sat inside teams where brilliant ideas died in Jira backlogs, and I’ve watched the opposite too: features rushed out fast, only to flop because they were poorly executed or customers didn’t care. Both hurt. Both waste time and burn cash.

Cameo Labs is my answer to that problem. I started it because I couldn’t stand watching founders lose momentum on ideas that should have changed their business and their lives. 

My mission is simple: help founders build the right product, fast enough to matter.

Since starting, I’ve led projects that:

- Shipped an EdTech platform that closed $3M in contracts in under five months.
- Migrated 1,000+ servers with zero downtime, saving $6M.
- Built an AI training platform that scaled to $1.5M ARR in 9 months.

I’m dorky about this stuff. I love the messy early product conversations, poking at assumptions, figuring out where speed really matters, and then helping a team pull off what feels impossible. That’s what gets me up in the morning.

If you’re a founder under pressure to ship something that can’t wait, I’d love to hear about it.

Chief Product Officer / Fractional Product Management

As the owner and founder of Cameo Labs, I am dedicated to helping CEOs and executives of $100M - $500M businesses in the Edtech, Fintech, and SaaS sectors achieve rapid and transformative innovation. 

At Cameo Labs, we understand that staying competitive in today’s market requires more than just innovative ideas—it demands strategic execution and a customer-centric approach.


I founded Cameo Labs to empower our clients by streamlining their product development processes, aligning their business strategies with technological advancements, and fostering a culture of continuous innovation.

We operate differently than most agencies. 

Our holistic approach combines technology, strategic foresight, and customer insights to create cohesive and impactful product visions. Our clients have experienced faster innovation cycles, improved product success rates, and higher customer satisfaction.

With a proven track record of success, we are committed to turning your challenges into opportunities for growth and success. Our goal is to be your trusted partner, helping you achieve your business vision and lead your industry.

Let’s Connect:
Discover how Cameo Labs can transform your business. Visit http://cameolabs.com or reach out to me directly to explore how we can work together to drive your business forward.

From Vision to Execution | Faster Products, Better Results, Bigger Impact | SaaS Companies | Strategic Software Partner

What fresh hell is this? 

I swear the point of AI was to make it so we didn’t have to work so hard but instead we’re using it to work even while we sleep? 

I actually really like the idea of seeing what my lucid dream brain could do with AI, but please don’t make it always be about work.

How is Sam Altman so sure he knows what people want?

Back in his YC days, he was famous for telling founders they didn’t know what their customers wanted until they actually talked to them. 

Now he’s out saying that cost isn’t the bottleneck, speed is.

Did he talk to anyone?

I work with founders and executives on AI strategy every week. Nobody is asking for more speed. 

Right now the biggest concern is accuracy. Can we trust this thing when it’s wrong so often? How do we build something reliable when the models hallucinate and we need human review for everything. 

Speed doesn’t matter if the answer is garbage.

You know what takes longer than waiting 10 seconds for a response? Spending 20 minutes fixing output that is screwed up.

Sam says people want results in 1/100th of the time even if it costs more. The people I talk to would pay more for results that are actually correct, even if they take longer.

But that’s OK because at this point, it’s pretty hard to take any claim from Sam Altman seriously anyway. 

What do you think? Do you think speed will be the competitive battlefield for LLM’s?

No alternative text description for this image

A couple of weeks ago Satya begged people to stop calling AI output slop.

So of course someone wrote a Chrome extension that replaces “Microsoft” with “Microslop” all over the web. 😂😂

This is what happens when billionaires ask the internet to be nicer.

The Streisand effect is real. You can’t tell people to stop criticizing something and expect them to comply. 
Especially when that something affects how they work, what they create, and whether they have jobs.

If you don’t want people calling your output slop, make better output. If you don’t want people questioning your technology, build technology worth trusting.

Asking for less criticism while scraping human knowledge, justifying mass layoffs, and rejecting intellectual property rights is wild.

graphical user interface, text, application, email

I Have always hated the annual review process. What a waste of time and effort for everyone involved. 

As a manager I was often asked to participate in the sham, and pretend that it was justification for the raise (or lack thereof) that has already been decided by upper management.

Futurist & Tech Investor 😇 | Keynote Speaker: AI, Blockchain & Emerging Tech | Author of The “Strategic Millionaire” and the upcoming “The Future of Wealth” | Helping Audiences Build Wealth Through Future Tech Investing

On the surface, this seems like a nasty side effect of Al.

But really people have been trying to get away from reading the Classics for as long as there have been Classics.

That's why Cliff Notes had a thriving business for so long. But not anymore.
RIP Cliff Notes.

With Cliff Notes, you knew you were cheating. You knew you weren't reading the real book.

Do you think people realize the same for this?
"Maximize your reading potential and avoid difficult language today" sure does sound catchy.

I don’t want crappy software. 
I don’t wanna build it and I sure as hell don’t wanna use it.

I saw this post in my feed. 
The OP's argument? 
Human developers and teams always shipped garbage anyway, so AI-generated garbage is fine. 

We have lived with bugginess, and random software issues forever, so why should LLM code be different. 

This logic is insane. 

"We have always been bad at this" isnt a defense. It isnt even true. 
Yes, a lot of software is really bad. A lot of teams are bad at building software. 
But that sure as hell doesnt make it ok. 

As a consumer, I will continue to demand better from the software that I use, and as a builder, I will continue to demand better from myself and my team. 

His last line is a doozy too. 
Builders who care about craft aren’t protecting their egos. They’re protecting their users from garbage software.​​​​​​​​​​​​​​​​

Do you still believe standards matter? Let’s talk.

~~~~~~~~~~~~~~~~~

Hi I’m Cameo, I help teams use AI without the garbage output.​​​​​​​​​​​​​​​​

There is no moat. 

Cursor just released their 2025 year-end data and what it shows is fascinating. This image is about AI coding tools specifically, but I think we are seeing it happen to SaaS too.

Here is what this image shows. When there is reasonable similarity between quality, everyone treats the tools like a commodity.
Look at the pattern. Claude 3.5 Sonnet dominates for a bit, then Claude 4 Opus launches and usage spikes. Then Gemini 2.5 Pro, and Claud 4.6 Sonnet. 

Each time something new drops, developers jump.
Nobody is loyal. They're chasing "good enough for the task" and switching the moment something newer shows up.

I think this is coming for all of SaaS. Not just the AI wrappers. All of it.
If someone can vibe code a good enough replacement for your product in a weekend, why are they paying you $300/month? The old moats were switching costs, integration complexity, and internal learning curves. 

But AI is flattening all of that. 
- Switching costs drop when an AI can migrate your data. 
- Learning curves disappear when the AI can generate exactly what you need.

So what's the compelling reason for customers to stick with your product?

It's never been more important that your moat is a product people actually love. One that truly meets their needs. Because if it doesn't, they're gone.

~~~~~~~~~~~~~~~

Hi, I am Cameo
I help founders figure out the right thing to build and then actually ship it.

Happy Groundhog Day!

It’s funny how the same predictions get made again and again.

Claude code has improved significantly in the last year. But I’m still not firing my software developers.

graphical user interface, text, application, chat or text message

It’s fun to watch Sam A out hustling to try and justify a $830B valuation.

I feel like we dont really need anything more testing us right now.

graphical user interface, text, application

Essentially the whole LLM discourse on LinkedIn right now. 

Says “agents” three times in a row and suddenly you are a CTO

Most companies are slow for one boring reason. 
Work spends too much time waiting for someone else.

Most organizations vastly underestimate how much time their people spend coordinating and shuffling work through their systems.
In fact a lot of my time in Agile transformation with fortune 500 companies was spent trying to help people spend less time orchestrating work.

Right now, a lot of software companies and development teams are trying to figure out how to make AI effective. But they have insidious data problems or serious, unresolved technical debt. 
If you try to solve all of that first, you are going to fall behind.

Stop aiming AI at the hardest problems first. 

Use AI to show you what is blocking your teams. Use AI to show you where work is getting stuck, where it is waiting on approvals, where it is bouncing between people, and where nobody owns the next step.

Then remove the orchestration tax you are paying every single day.
Organizations who do that can feel the impact before they ever ship a single “AI feature.”

~~~~~~~~~~~~~~~~~~~~~~

Hi, 👏 I am Cameo I help companies turn AI experiments into actual results.

RIP to VR/XR! 

Not so fast

I have long been bullish on virtual and augmented/extended reality, not the cartoon stuff that was in the horizon workrooms, but the potential.

The headlines want you to believe that Meta is giving up. That their recent laydowns and shutdowns mean they are admitting failure.

But reality labs only cut 10%. That leaves them with at least 13,000 people still working on extended reality.

I know a lot of people believe that in the future, human AI interaction will move entirely to voice, but I don’t buy it.

To be truly effective as a tool, AI needs to be able to see things in real time. As it is, I’m constantly feeding screenshots and photos into LLMsto shorten how much explaining I need to do. 

A picture is worth 1000 words, right?

And I see a ton of value in being able to overlay information into our view, especially if I’m trying to learn something new. That’s the promise of XR.

And yes, I’m sure Meta squandered a lot of that roughly $70B plus, but they also laid the groundwork for a lot of what comes after.

So what do you think is next for VR/XR?

The contrast between two images:

On the left…
I am sure you have seen people’s posts about their Lovable or Replit year in review.... Bragging about how many lines of code they generated. 

And of course, my first question is, when did we decide that lines of code is a useful metric at all? And who decided it? 
People don’t buy code. They buy products and functionally that solve problems. (To be fair, the US government often pays based on lines of code, but they’re infamous for making really crappy software)
But the implication is that those millions of lines of code are being generated to build software products that people want and need... Right? 

And then the image on the right is from a substack article that’s been making the rounds by Mike Judge. (Link in comments) 

His premise is that if vibe coding and AI enable development is actually having the enormous impact that everyone claims, we should see a huge explosion of shovelware, and even as we enter 2026, that has not happened. 

Personally, I think the truth is somewhere in the middle. 

Over the last year I’ve built multiple applications in Replit and Lovable that I am using every day in my business and in my life. 

But I’m not trying to spin them up as businesses. I’m using these tools to solve inconveniences in my life with apps that I don’t need to be secure, scalable or any of the rest of the nonfunctional attributes that AI development fails at. 

I'm curious, have you read Mike's shovelware post? 
What do you think is happening?

~~~~~~~~~~~~~~~~~~
Hi I am Cameo
If your team is generating millions of lines of code but you're not seeing business impact, let's talk about what's actually possible versus what just makes for impressive demos.

The most important skill you need for 2026 is vibecoding.

This is especially true if you work in technology but aren’t a software developer. If you’re in leadership, marketing, sales, or HR, you need to understand what’s possible with these tools. 

More importantly, you should be experimenting with them and using them to build small things day to day.

Vibecoding was a game changer for me and my business in 2025. Don’t get left behind in 2026.

As part of Utah Tech Week, Krystee Miller and I are hosting a workshop on how to vibecode with Lovable. 
In 90 minutes, we’ll show you how to shape and refine an idea, determine what it needs and how it should work, and then prompt it into existence using only your voice.

The event is free. Lunch provided. Link to register below.

Guys, 
I have discovered OpenAIs secret plans to make ChatGPT profitable. 







*

"Stop being critical of our AI"

Over the last few days, a couple of billionaires came online to beg us to stop saying mean things about their LLMs.

First Satya asked us to stop calling it slop.
Then Jensen said being critical of AI isn't helpful.

Seriously, what kind of joke is this?

Over the last three years, we've watched these companies pour hundreds of billions of dollars into this technology. We've watched them use it to justify some of the largest layoffs in recent history. We've watched them scrape the entirety of human knowledge and completely reject concepts like IP and copyright that are supposedly the bedrock of capitalism.

And we're not supposed to be critical?

I use AI heavily. At Cameo Labs we use it effectively with our clients for developing software faster. I've vibed coded multiple applications for both my business and personal life. I've used to to create interesting and creative things that I think are pretty cool.

AND I'll still say a lot of current AI output is mid at best. The fact that they don't want us to criticize it tells you everything.
As members of society, we get to criticize things that seem detrimental to our collective success. 

You can be pro-AI and still demand quality and accountability. Using these tools every day doesn't mean accepting garbage or accepting evil described as progress. (looking at you Grok) 
That's not being a hater. That's having standards.

The difference between slop and substance isn't the tool. It's knowing what you're building and refusing to ship or accept crap.

~~~~~~~~~~~~~~~~~~~~

Are you curious what vibe coding is and want to get your feet wet and still build something cool? I'm hosting a vibe coding workshop at Utah Tech Week 
Link in the comments.

Listen, I know it’s an awkward conversation, but we need to talk about pirates booty

This is a super popular snack in Utah, and is often served in corporate environments but I cannot see this package without instantly thinking about the Beastie Boys song, Professor Booty, from their seminal 1992 album Check Your Head. 

I quote: 
"Professor, what's another word for 'pirate treasure'?"
"Well, I think it's 'booty-booty-booty, that's what it is"

Now the song lyrics are NSFW, so having this song running through my head in a professional environment it’s sometimes problematic.
Plus I am trying to work, not sing Beastie Boys songs. 


Surely I am not the only one who has this problem?

Vibe coding works until it doesn't

Turns out the edge cases matter. They always do.

Your users find the problems you didn't think to look for. And they're not happy about it.
Test the weird inputs. Try to break your own code. Think about what happens when two things occur simultaneously, when files are too big, when numbers are negative. When security actually matters. 

Do it before your customers do.

Expert catalyst in driving Innovation from 0 to 1, excelling in building innovation teams, launching new products, embracing emerging tech, and a passionate couture design enthusiast

When leaders face their most complex, mission-critical challenges, the ones that keep them up at night, they call me.

I'm a strategic advisor and execution partner who transforms emerging technologies into measurable business results. For over two decades, I've been the trusted right-hand to leaders across global enterprises and fast-growing ventures alike navigating their most complicated transformations, from AI adoption to digital overhauls to innovation at scale — unlocking growth, reducing costs, and building lasting competitive advantage.

I don't just give advice, I roll up my sleeves and build alongside you. Whether it's leading AI adoption strategies, driving digital transformations across multiple brands, or tackling the high-stakes, multi-stakeholder initiatives where failure isn't an option and collaboration is key. With deep technical expertise and business acumen, I bridge the gap between strategy and action.

My work spans financial services, consumer goods, technology, and more. I'm recognized as a thought leader and regularly advise industry organizations and universities on emerging technology trends and implementation strategies. 

I'm passionate about helping leaders navigate what's next. So, let's talk about what keeps you up at night.

Darlene Newman

Most companies don’t have an AI strategy problem. They have a strategy definition problem.

What’s being called an “AI strategy” is usually a collection of tools, pilots, and platforms.
Useful… but not strategic.

A strategy only matters if it changes how the organization works.

Traditional AI struggled to deliver at scale not because it failed, but because it was hard... narrow models embedded at specific decision points, expensive to integrate, and difficult to reuse.

The value was real, but limited.

Generative AI changes the equation.
Not because it’s easier, but because it has the potential to change how work gets done across tasks, roles, and workflows.

And that’s exactly why most AI strategies still won’t stick.

If organizations struggled to understand how traditional models worked and what had to change around them… that challenge hasn’t gone away with LLMs.
AI only creates durable value when the strategy reaches past the tools and into how the organization actually operates.

That’s the difference between an AI plan and an AI strategy.

I’ve laid out what an AI strategy that sticks actually requires.

Strategic partner for leaders' most complex challenges | AI + Innovation + Digital Transformation | From strategy through execution

AI is Not the Strategy - Ivy Captech Advisors

The consulting world is splitting into two camps on enterprise AI… and both are setting their clients up for problems.

In one corner are the transformation evangelists. McKinsey tells us the “weights have shifted toward value” and organizations should pursue company-wide AI transformation. BCG’s model isn’t much different.

The message is, go big or go home.

In the other corner, a counter-narrative is forming. Unisys just released its 2026 IT Insights report arguing that “focused AI deployments will outpace transformation projects.”

The message here? Stop boiling the ocean. Focus on smaller, nimbler, less threatening AI tasks.

Here’s the problem... both camps are right about the other’s blind spots, and wrong about their own.

The big bang AI playbook assumes organizations have things they almost never do:

🔹 Data that actually makes sense across the enterprise. Not just “clean,” but shared definitions, ownership, and context. Most organizations don’t have this
🔹 Leadership alignment that doesn’t evaporate under pressure. Company-wide AI requires sponsors who can tolerate ambiguity. Fear, not feasibility, is what usually kills these efforts
🔹 The stomach for change. Large-scale AI rewires how people work. That means prolonged guidance, friction, and discomfort... realities most organizations underestimate or avoid

Transformation is awesome in a board deck. But anyone who’s had to innovate from a Mckinsey deck will know the reality is very different.

The “start small” advice is directionally right, but incomplete. The problem isn’t starting small often stops there.

When organizations pursue isolated, task-level deployments they often create:

🔹 AI sprawl. Dozens of point solutions across functions, each with its own platform, data logic, governance approach, and metrics
🔹 Compounding technical debt. Every pilot makes independent choices about models, prompts, controls, and oversight, turning future integration into a nightmare
🔹 Lost learning. Teams rediscover the same lessons about data quality, adoption, and edge cases because no one is synthesizing insights across projects

This is how you end up with “successful” AI pilots that never add up to strategic advantage.

What actually works: federated enablement

The answer isn’t transformation or incrementalism. It’s building the connective tissue that lets focused deployments compound.

➡️ Start with the operating model, not the use case. Who owns governance?  How is data accessed? Who defines the shared context? 
➡️ Enable the edges, govern the center. Shared guardrails... models, security, evaluation, measurement, with decentralized execution.
➡️ Create an ecosystem that supports experimentation and production.
➡️ Measure the portfolio, not just the project. Are deployments getting faster? Are components being reused? Are second wave initiatives outperforming the first?

It’s not big or small. It’s whether your AI efforts can compound.

Navigating the Future of Tech and Business

We hear a lot about “context” in AI. But very few people explain what that actually means… or what it takes to make it real.

I came across an article from Enterprise Knowledge that puts a precise name on the issue: the missing semantic layer. It also helps explain why so many AI projects fail.

Without it, AI breaks down in predictable ways:

1️⃣ AI misinterprets organizational context

AI can’t distinguish between what your company means and what your data says. When Marketing’s “customer,” Finance’s “client,” and Legal’s “account holder” don’t align, AI stitches together mismatched concepts. The output sounds confident… but it’s often wrong.

2️⃣ AI can’t connect knowledge across silos

The average enterprise has data spread across 5–30+ systems. AI can read a contract, but it can’t reliably tell that the supplier named there is the same entity tracked in procurement, finance, and risk… unless those relationships are explicitly defined.

3️⃣ AI becomes a black box nobody trusts

If AI can’t explain its reasoning, leaders won’t trust it, compliance can’t audit it, and teams can’t fix it when it breaks.

So what is a semantic layer?

It’s the layer that gives AI a shared understanding of your business... what terms mean, how things are organized, how concepts relate, what context matters, and how it all connects.

Think of it like onboarding a new analyst. They may have access to every system, but without business context, all they see are fields like CUST_ID and REV_AMT_NET_ADJ. They don’t know the rules behind the data, or what “active client” really means.

Here’s what that looks like in Contract Lifecycle Management, which I’m working through with a client right now:

🔹 Business glossary (shared definitions)
What does “contract value” mean? TCV or ACV? Which date is “effective”?

🔹Taxonomy (how things are organized)
How are contract types organized? Master Agreement → SOW → Amendment. Status from Draft to Active to Expiring.

🔹 Ontology (how concepts relate)
Contracts have parties. Contracts contain clauses. Clauses create obligations. Obligations have owners, deadlines, and risk. Contracts relate to other contracts through renewals and amendments.

🔹 Metadata (operational context)
Which version is authoritative? Who owns the contract? Which entity is the counterparty? What jurisdiction applies?

🔹 Knowledge graph (the connective tissue)
A queryable map linking suppliers, contracts, clauses, obligations, risk, and systems… so AI can answer cross-cutting questions without guessing

And, where should this Semantic Layer live?

Not in slide decks. Not in someone's head. Not rebuilt inside every application.

The semantic layer sits above systems of record (CLM, ERP, CRM) and below analytics, agents, and copilots. Source systems hold the data. The semantic layer governs what it means.

In the end... it’s about recognizing that AI without organizational context is just pattern-matching across disconnected information.

Article in comments

Randy Bean and Tom Davenport just published their 15th annual AI & Data Leadership survey in HBR. This is the 15th year of this benchmark… 110 Fortune 1000 companies, 96% C-level respondents, heavily weighted toward financial services, healthcare, and retail.

At first glance, it reads like a victory lap…

🔹 99% say data & AI are a top priority
🔹 91% are increasing investment
🔹 AI in production jumped from 4.7% to 39.1% in two years
🔹 Only 6.4% remain in pure experimentation
🔹 90% of companies now have a Chief Data Officer
🔹 38.5% have appointed a Chief AI Officer

The authors are calling this the "Golden Age of AI."

So, we’re finally winning, right?

Not quite.

Buried in the survey data… 93% say culture and change management remain the greatest barrier to AI adoption.

Not technology. People

And not solved by just placing another executive in charge unless they have a mission beyond data and AI

And despite record investment and executive proliferation, only 54% report high or significant business value from AI.

Half the enterprise is still struggling to translate ambition into outcomes.
There’s another signal hiding in plain sight:

No one agrees where AI leadership should even sit.

CAIOs report to:
🔹 Technology (34%)
🔹 Data (30%)
🔹 Business leadership (27%)
🔹 Transformation (9%)

The authors suggest this confusion may be contributing to the perception that AI isn’t delivering.

Where they sit, doesn’t matter…. Here's what I think we're getting wrong.

What enterprises are calling "AI" right now isn't traditional AI. (sorry folks, you’re not building AI, you’re using a model to run workflows).

It's not data science. It's agentic capability…  LLMs, low-code tools, workflow automation, embedded copilots… (I know.. it’s disappointing).

It's digital transformation by another name.

The leader who drives this doesn't need to be an expert at models or statistics. They need to understand three things:

--> Data (what's available, how it flows, and what business context need to be build around it to make AI understand it)
--> Digital (process design, user experience, integration)
--> AI/agentic tools (what's possible, how to orchestrate, what type of model… LLM, deterministic, do I have to go as far as machine learning)

But more than anything, they need to own enablement.

It’s solved by:

🔹 Upskilling people where work actually happens and fluency in how to define the path for implenetation
🔹 Designing self-service paths that reduce dependency on central teams, but governed and secure
🔹 Helping employees learn how to safely use, adapt, and trust these tools

The authors suggest closer collaboration with CHROs and stronger change capabilities inside CDO teams. That’s directionally right.

But let’s be honest about the root issue.

It’s an enablement problem. Not just an upskilling or executive assignment issue.

2026: AI & Data Leadership Benchmark Survey

By 2030, AI won’t just support the business model… it will be the business model.

That's the core argument of IBM's "The Enterprise in 2030"... not a prediction about AI capabilities, but how organizations will reorganize once intelligence is embedded everywhere.

Here's the shortened version of their top 5 predictions… in case you don't have time for all 63 pages.

1️⃣ Competitive pressure will make big bets unavoidable

What does that mean? AI flattens the playing field. Startups can now run at enterprise scale while moving faster than incumbents ever could. Competition also shifts from isolated firms to ecosystems of partners that build, integrate, and go to market together.

IBM says:

🔹 Make larger, asymmetric bets early
🔹 Stress-test business models through rapid experimentation, not long-range plans

2️⃣ Productivity gains are just the opening act

IBM describes two phases. 

--> Phase 1: AI eliminates waste, accelerates work, amplifies human capability. This is where we are now. 
--> Phase 2: leverage freed-up resources to reimagine entire verticals and business models.

IBM says:

🔹 Set explicit productivity targets tied to AI
🔹 Run recurring "efficiency sprints" on core workflows
🔹 Redeploy saved capacity into new products, services, and capabilities

3️⃣ The best AI will be one-of-a-kind… not generic

Access to powerful models will be universal. Differentiation comes from AI tailored to proprietary data, workflows, and decision logic, using portfolios of large and small models that work together.

IBM says:

🔹 Identify capabilities that truly differentiate you
🔹 Build models around them
🔹 Create an orchestration layer so models, agents, and systems can evolve without lock-in

4️⃣ AI won’t do all your thinking… but it will change whose thinking matters

As more execution and knowledge work shifts to agents, human value moves up the stack. The most important skills won’t be technical, they’re judgment, decision-making, strategy, and collaboration.

IBM says:

🔹 Redesign roles around orchestration and oversight, not task execution
🔹 Embed learning into workflows
🔹 Hire for problem-solving and systems thinking, not just expertise (The business-to-tech bridge role will be critical)

5️⃣ Leadership and governance will be redefined

Traditional, human-based decision models won't keep up. As AI participates in decisions, confidence shifts from hierarchy towards organizational intelligence… clear accountability, explicit decision rights, governance that enables speed while not losing control.

IBM says:

🔹 Clarify where AI can decide, recommend, or escalate
🔹 Build governance that scales with automation
🔹 Prepare leaders and boards to work with AI-assisted decisions, not around them

IBM isn’t describing better AI capabilities for 2023. They’re describing a different kind of enterprise… one designed to think, learn, and adapt continuously.

The report is worth the read… even all 63 pages.

IBM: The Enterprise in 2030

When “Wartime Speed” meets legacy system reality.... it’s not that easy.

The Department of Defense just released a new AI strategy, and it makes one thing very clear… speed outweighs perfection.

The memo states that the risks of not moving fast  enogh outweigh the risks of imperfect alignment. It mandates 30-days to share data and deploy models, and empowers leadership to waive deployment blockers for speed.

I get it. 

Perfection is the enemy of progress. Innovation requires testing, learning, and accepting some failure.

But speed on paper and speed in reality are not the same thing, especially inside legacy, mission-critical systems.

Strip out the politics and you’re left with a simple question: can a legacy organization actually move at AI speed?

Having led innovation inside critical financial infrastructure that underpins the markets and processes trillions of dollars in transactions every day, I can say this plainly: mandating speed, and measuring success by speed alone, doesn’t create speed.

Because, it's not just processes, data ownership and deployment approvals slowing things down.

🔹 It's data and system logic that only a handful of people understand. And if it’s a Cobal system, very few people in this work even know how to work it
🔹 Integrations that don't exist yet
🔹 Agent capabilities that can't even connect to legacy systems built decades ago
🔹 Ownership buried across teams that don't talk to each other
🔹 Dependencies that were never documented because the person who knew them retired five years ago

Mandates can't fix this.
Those are execution realities every large, legacy, mission-critical organization should pay attention to… far beyond defense.

If the Defense Department, or any such legacy organization, wants to move at AI speed, mandates won’t get it there. Here's what will:

1️⃣ Design around legacy constraints instead of pretending they aren’t there. You can’t reimagine your way past systems still running live operations; speed comes from defining where AI can safely plug in and where it can’t.

2️⃣ Shared meaning matters more than shared access
Access to data isn't the bottleneck, common understanding is. If teams don’t agree on what data means, AI can’t help.


3️⃣ Optimize for decision velocity, not deployment velocity
Shipping fast and operating fast are different things. Value comes from better decisions, not faster releases.

4️⃣ Test aggressively, promote carefully
Failure accelerates learning in test environments, but not in production. Production failures... it's a $50K per hour problem.

5️⃣ Build observability before autonomy
You can’t scale agents you can’t monitor, explain, or audit. Autonomy without visibility isn’t speed... it’s manafement chaos.

None of this is ideological. It’s the physics of legacy organizations.

Wartime AI speed has to be matched by legacy modernization, because how fast you can deploy AI is entirely dependent on whether your systems can absorb that speed without breaking.

AI Strategy for the Department of Defense

If your Agent implementation strategy starts with a full process map, you’re already doing it wrong.

This contract lifecycle flowchart... it's useful. It shows where work happens, where decisions get made, where exceptions route.

But it's not an agent design. It's a map of where agents might eventually live.

And that's the problem. Most teams are still building AI the way they built traditional software.

1️⃣ Pick a use case
2️⃣ Design the end-to-end process flow
3️⃣ Identify rules, exceptions, and approvals
4️⃣ Design the interface and wireframes
5️⃣ Build integrations and orchestration
6️⃣ Then apply AI to a specific decision point

That approach worked because AI wasn’t the organizing principle. It was process optimization… not process transformation

AI Agents change that entirely.

Agents are narrow by design. And should be.

They handle one task, one decision, or one action… not an entire workflow.

Why? Because they fail when they’re designed to take on too much. Errors start to compound. Context gets polluted. A hallucination in step 3 cascades through steps 4, 5, and 6. 

And now you've built something you can't debug, can't trust, and don’t know how to fix.

Workflows tell you where work happens. Agents are built to perform specific tasks inside those workflows… reliably and on their own.

This is where I see so many AI initiatives break.

Most teams define the entire flow as the use case. Then they try to build agents to support it in a nice package to show off to leadership.

Why not? That's how you tell the story of HUGE transformation.

Unfortunately, that concept of “transformation at once” forces them to design:

🔹 Interfaces before they know whether the agents actually work
🔹 Integrations before they know what data is truly required (and I’m talking context)
🔹 End-to-end behavior before they’ve proven any part of it

Agents are emerging tech. And emerging tech needs to be treated like R&D… not big-bang delivery, but small, incremental testing and learning that compounds into transformation over time.

Agent-first design starts by pulling individual tasks out of the workflow and treating each one as its own experiment.

Define the inputs.
Define the output.

And if those inputs don't exist cleanly? Congratulations, you've found your first problem to solve. 

That's not a failure. That's the work most teams skip and then wonder why their agents underperform.

Then test whether the agent can perform the task reliably

🔹 without a full UI
🔹 without full orchestration
🔹 without pretending you already know how it will work

Run it. Measure it. Break it. Learn from it. Not perfectly… but in production.

Only after you understand what the agent can actually do do you start pulling those capabilities back into a broader flow.

No, the flowchart isn't wrong. It's just not how you define the execution path.

Boston Consulting Group's AI Radar 2026, a global survey of 2,300+ executives including 600+ CEOs, makes an interesting claim: as AI investments surge, CEOs are taking direct ownership of AI decisions.

But the data in the report tells a more complicated story.

On the surface, this sounds like progress. AI is finally a CEO-level priority. 74% now see themselves as the primary AI decision-maker.

But BCG is also explicit that AI success requires organizational transformation across strategy, operations, culture, risk, talent, and data. That’s enterprise-wide work.

So which is it? A distributed organizational capability… or one person's call?
BCG segments CEOs into three groups:

🔹 Followers (15%) - cautious, pilot-heavy, high anxiety
🔹 Pragmatists (70%) - advancing "with the market," not ahead of it
🔹 Trailblazers (15%) - AI as top priority, large-scale deployment, ~75% of workforce upskilled

But flip the lens…

🔹 85% of CEOs are not Trailblazers
🔹 74% say they personally own AI decisions
🔹 50% believe their job is on the line if AI doesn’t deliver

That's not organizational readiness. That's concentrated accountability without distributed capability.

When the CEO becomes the de facto Chief AI Officer, three things tend to happen:

1️⃣ Decisions outpace execution capacity
Investment can be greenlit instantly, which is great for momentum… enterprise level capability unfortunately, cannot. Skills, workflows, data readiness, and ecosystem integration take time.

Without workforce enablement, execution lags... regardless of who makes the call.

2️⃣ The confidence gap becomes structural
BCG’s data shows confidence drops sharply from CEOs to non-C-suite leaders because the two groups are responding to different realities.

--> CEOs see strategy: investment cases, vendor promises, board pressure, projected ROI.
--> Executors see constraints: fragmented data, brittle systems, unclear authority, and workflows that aren’t AI-ready.

3️⃣ Accountability gets personalized, not institutionalized
If AI fails, it's the CEO's failure. If it succeeds, did the organization build a durable capability… or just get lucky on a few bets?

Trailblazer CEOs, the 15% that are differentiated, aren’t successful because they personally own AI decision making.

They’re successful because they’ve ensured the unglamorous work first:

☑️ Data, governance, and operating model foundations
☑️ Workforce-wide upskilling
☑️ Streamlined path from experimentation, to pilot, to scale

CEO ownership is the result of readiness… not the cause of it.

BCG’s implicit story says… CEO owns AI → drives investment → organization follows
But the actual model that works… Organizational readiness → enables AI adoption → CEO provides air cover

One concentrates risk.
The other distributes capability.

So if the CEO in your organizations owns decision making… ask yourself “why”.

Because the organization is ready?
Or because no one else can?

Report in comments

The companies getting real value from AI aren’t better at picking use cases. They’re better at redesigning how work gets done.

That theme runs through the World Economic Forum’s latest report, Proof over Promise: Insights on Real-World AI Adoption from 2025 MINDS Organizations, released with Accenture.

MINDS (Meaningful, Intelligent, Novel, Deployable Solutions) is WEF’s global program moving AI past hype and pilots, spotlighting deployed, in-production use cases with measurable impact. The report draws on hundreds of applications, deep interviews with the 2025 cohort, and quantified economic, environmental, and social outcomes.

Rather than asking what's possible with AI, the report focuses on what's actually working… and why.

Organizations seeing the biggest impact aren't asking where AI fits best. They're asking how AI can reimagine how work gets done.

A subtle but critical distinction… and the foundation for four themes that kept repeating:

1️⃣ AI is treated as an operating capability, not a project

In one of the strongest examples, AI wasn't deployed to "support" an existing process, the process itself was reimagined.

Rather than applying AI separately to forecasting, trading, grid regulation, or settlement, State Grid Corp of China built a single orchestration layer that coordinates all four. AI didn't optimize steps in isolation. It managed the system as a whole, with humans overseeing critical decisions.

2️⃣ Human–AI collaboration is explicitly designed

The report identifies three governance patterns for AI… full autonomy, bounded autonomy, and human-governed autonomy.

That's not new of course. What matters is that leaders were choosing intentionally, based on risk and context.

The organizations scaling AI aren't debating "human in the loop" as a theory. They're engineering it into the process.

3️⃣ Contextual understanding is mission critical

Despite all the generative AI enthusiasm, data quality and contextual understanding remain the most common barriers. Successful organizations close the gap by grounding AI in institutional knowledge.

SAP paired an AI copilot with 200K pages of proprietary documentation and best practices… expertise that typically is tribal knowledge.

The breakthrough wasn't the AI. It was making context available.

4️⃣ Platforms beat point solutions

Nearly every scalable example converges on the same idea… AI needs a unified platform.

ICBC's financial AI platform, Siemens' industrial AI stack, and State Grid's energy orchestration system all act as foundations… allowing new agents, models, and use cases to be deployed in minutes instead of built from scratch.
That's the difference between scaling AI and repeatedly reinventing it.


What was so interesting about this report, wasn’t just in reading the use cases, but was in realizing that AI success isn't about finding the right use case.

It's about building the conditions where many use cases can emerge, evolve, and compound over time.

taplio

The Battle

Cameo Doran

Darlene Newman

Followers

Posts

Engagements

Likes

Performance ScoreCameo Doran's LinkedIn Performance

Performance ScoreDarlene Newman's LinkedIn Performance

Ready to grow your LinkedIn presence?

taplio

The Battle

Cameo Doran

Darlene Newman

Followers

Posts

Engagements

Likes

Performance Score.css-goub6s{font-size:var(--chakra-fontSizes-md);font-weight:500;color:var(--chakra-colors-gray-600);display:block;margin-top:var(--chakra-space-1);}Cameo Doran's LinkedIn Performance

Performance ScoreDarlene Newman's LinkedIn Performance

Ready to grow your LinkedIn presence?

Performance ScoreCameo Doran's LinkedIn Performance