The AI Economy: Five Stages to Achieving AI Maturity

Welcome to "The AI Economy," a weekly newsletter by Ken Yeung on how AI is influencing business, work, society, and technology. Subscribe now to stay ahead with expert insights and curated updates—delivered straight to your inbox.

For this week’s issue of “The AI Economy,” take a look at a new study from Asana and Anthropic that offers a roadmap for organizations looking to make artificial intelligence a core part of their business strategy.

Plus, here’s what we should expect and what we shouldn’t hold our collective breaths for at Apple’s WWDC conference next week, and learn about the new program from Cohere that supports early-stage startups.

From AI Skeptic to AI Maturity

It seems that every week, new research study findings pop up about the state of AI, especially in how it’s influencing the world of work. Many of these have the same core message—workers are adopting the technology, but business leaders struggle to develop effective use cases—though the numbers behind the investigation vary.

However, a report from Asana and Anthropic entitled “The State of AI at Work” caught my attention. In addition to its statistical findings, the team detailed a classification schema to help companies understand their maturity level with AI. It outlines five factors that go into evolving from one stage to the next.

“The report is the first to classify organizations according to a five-stage AI maturity model, from immature (Stage 1) to mature (Stage 5) implementations,” Dr. Rebecca Hinds, Asana’s Work Innovation Lab chief, explains to me. “The study also outlines the steps organizations need to take to achieve mature AI implementations, known as the ‘5Cs’: Comprehension, Concerns, Collaboration, Context, and Calibration.”

More than 5,000 knowledge workers in the U.S. and the UK were polled for this study, which aimed to educate decision-making executives on how to implement effective AI strategies and use the technology to boost productivity and collaboration. Unsurprisingly, more than half of respondents fall into the first two stages of organizational AI maturity, namely AI Skepticism and AI Activation.

What Are the Five Stages?

Asana and Anthropic’s 2024 “State of AI at Work” report highlights the five stages companies go through to achieve AI maturity. Image credit: Asana/Anthropic

The first stage is AI Skepticism, in which companies begin to recognize AI’s potential but don’t know enough to do anything with it. It’s followed by AI Activation, a stage when companies start testing the waters with pilot projects. It’s about gaining hands-on experience through small-scale experiments.

The third stage, AI Experimentation, is when larger AI initiatives are formed, and business leaders start to consider ways to incorporate the technology throughout their organization. Things begin to grow in scale in the penultimate stage, AI Scaling, with AI becoming part of company workflows and decision-making processes. Finally, with AI Maturity, organizations have now made AI part of their DNA, leveraging it to drive transformative results and using it strategically to align technology with organizational goals.

Evolving Through the Five “Cs”

Asana and Anthropic’s 2024 “State of AI at Work” report details the factors companies must achieve in order to move between the different stages of AI maturity. Image credit: Asana/Anthropic

To move from one stage to the next, Asana and Anthropic propose the five “Cs.” They’re the same for each stage, but like leveling up a character in a video game, the same must be done inside an organization. The “Cs” are comprehension, concerns, collaboration, context and calibration.

As you’d expect, for a company starting out with AI, their “score” across the five “Cs” will be low, but would be the reverse for those that have achieved AI maturity.

“In the context of the Collaboration ‘C’ (how humans collaborate with AI), we find that employees at mature AI organizations are 1.5 times more likely to view AI as a teammate. This shift in perspective leads to increased AI usage, enthusiasm, and productivity gains,” Dr. Hinds says. “Regarding the Context ‘C,’ organizations with mature AI implementations are more likely to have AI policies and guidelines in place, providing a clear context for AI usage. Related to the Calibration ‘C,’ while only 17 percent of employees at Stage 1 companies report that their organizations collect feedback on AI usage, this jumps to 91 percent at Stage 5 companies, highlighting the importance of calibration in mature AI implementations.”

WWDC 2024: Apple Jumps into the AI Fray

Apple’s time in the spotlight is finally here. All eyes are on the tech maker to see how it’ll respond to all the AI innovations from Amazon, Microsoft, Google, and OpenAI. Is the iPhone maker really falling behind the rest of the industry in adopting AI?

Bloomberg’s Mark Gurman reveals that Apple will introduce Apple Intelligence—or AI, get it? Nevermind. This system will support iOS and MacOS devices and be powered by OpenAI. This development is notable because there had been reports that Apple was thinking about having its iPhone be powered by Google Gemini. In addition, Microsoft might be nursing some hurt feelings with OpenAI’s Apple partnership.

Apple Intelligence will also work in the cloud and on-device, with the former using OpenAI and the latter using Apple’s in-house developed models.

While Microsoft has long abandoned Cortana (RIP) and Google has abandoned Google Assistant in favor of Gemini, Apple may not be throwing the towel on Siri. The original virtual assistant may be viewed as a forgotten part of the iPhone, but now it’s expected that Apple will infuse Siri with Gen AI to make it more conversational and versatile.

It’s also been suggested that Apple will roll out standard gen AI features such as voice transcription and image generation. In fact, Gurman reported that iOS users could soon have AI create custom emojis for them.

Will Apple reveal its own LLM or other AI developer tools? I’d say at least for now, the answer is no. The company is spending time working on ensuring AI isn’t added piecemeal to its products, instead applying perhaps that Jobsian-like philosophy around design to ensure consumers have the best experience possible and that it feels natural.

Apple’s keynote on June 10 will be something I won’t want to miss because it feels like the company is turning a page in its history. Everyone’s waiting to see how it’ll apply its legendary product magic to AI and hope that while it hasn’t really tipped its hat, Apple isn’t cowering in the corner and can still innovate.

Cohere Moves to Support Early-Stage Startups

Cohere AI launches a program to support early-stage startups. Image credit: Cohere

Cohere, a provider of enterprise-grade AI solutions, has launched a program to support early-stage startups. The Command R+ model maker says it will back “ambitious” entrepreneurs “looking to make the most out of AI to successfully scale their businesses.” Accepted startups with Series B funding or below will receive access to Cohere’s models at a 25 percent discount for an entire year.

Programs of this kind can help draw in new users and also inform on what models these companies will need as they evolve. This isn’t the first time Cohere has debuted this type of initiative—it says the current Startup Program is an updated version of a beta program it had previously.

“We wanted to make sure early-stage startups don’t get left behind as larger companies embrace AI to gain a competitive advantage,” Tatiana Shabanova, the leader of Cohere’s go-to-market team running this program, shares. “We received a lot of positive feedback and interest from customers after releasing our latest Command R series of enterprise-grade frontier AI models—which are highly efficient, multilingual, and excel at business-critical use cases like retrieval-augmented generation (RAG) for strong accuracy and Tool Use to automate complex tasks. This interest from customers inspired us to create the startups program to ensure companies of all sizes can access state-of-the-art AI technology at an affordable rate to scale their applications into production.”

Cohere is looking for companies developing innovative AI-powered applications or products to improve and streamline business operations. Applications will be reviewed on a case-by-case basis, and Cohere wants to support as many startups as possible in order to “demonstrate the practical benefits of AI for businesses today.”

But what reason do founders have to participate when other AI providers, such as Amazon, offer perhaps more enticing benefits? Shabanova pushes back and says Cohere solves two large barriers to entry for companies: cost and data security. “Cohere’s AI technology is designed with security and privacy at the core to ensure our customers’ data is protected. We have built a trusted platform by working with some of the world’s largest enterprises like Oracle, which we intend to use as a blueprint to work with companies of all sizes.”

She further elaborates, “By leveraging Cohere’s frontier AI models, participants can build powerful RAG systems. Oftentimes, startups working on RAG applications have to use models from different providers, but Cohere offers a one-stop shop to accelerate deployment. This is part of a wider effort to support the global business community in accelerating adoption of AI technology at scale to boost productivity and efficiency.”

Today’s Visual Snapshot

Infographic from Madrona Ventures detailing emerging AI Agents infrastructure. Image credit: Madrona Ventures

AI agents are becoming more commonplace, seemingly becoming the popular use case for gen AI and replacing chatbots. More tech companies are releasing their version of AI agents to help their customers manage internal workflows and processes and also customer service issues. They can be useful in helping us code, search for information, complete a spreadsheet or generate an image.

But what goes into developing and maintaining these autonomous actors so that they operate effectively and produce the most accurate results possible?

Jon Turow, a partner with Madrona Ventures and former Head of Product for Computer Vision at AWS, published the above infographic this week. It details the infrastructure needed to support AI agents.

“[A]gents today have lots of limitations,” he writes in a blog post. “They are often wrong. They need to be managed. Running too many of them has implications for bandwidth, cost, latency, and user experience. And developers are still learning how to use them effectively. But readers would be right to notice that those limitations echo complaints about foundation models themselves. Techniques like validation, voting, and model ensembles reinforce for AI agents what recent history has shown for gen AI overall: developers are counting on rapid science and engineering improvements and building with a future state in mind. They are speeding across the half-finished bridge…under the assumption it will be finished rapidly.”

Because we’re still early in the evolution of AI agents, this infographic illustrates the infrastructure needed today—who knows what it’ll be next year? But Turow believes that in the coming months, “as use cases solidify and design patterns improve,” the landscape will change. “The most useful infrastructure primitives in the near term are going to be the ones that meet developers where they are and let them build hand-crafted agent networks they control,” he pontificates.

The investor reveals key themes he sees emerging from agent-specific developer tools, agents as a service, browser infrastructure, personalized memory, authentication for agents, and a so-called “Vercel for agents” in which a distributed system exists to seamlessly manage, orchestrate and scale agent hosting.

Quote This

The future is really about personalized LLMs. I will have multiple variants of my LLMs. Every enterprise will also have their own LLM as well. That’s the future. I do not think all of us will share the exact same LLM.

— Zoom Chief Executive Eric Yuan on Decoder with Nilay Patel podcast explaining he believes many LLMs are similar and have no big difference. In the future, everyone having their own LLM makes sense because it will be a model that can represent you anytime and really understand you. (The Verge)

This Week’s AI News

🏭 Industry Insights

Digital Twin versus simulation: What are the key differences? (Digital Twin Insider)
OpenAI’s ChatGPT, Anthropic’s Claude and Perplexity all went down on the same day (TechCrunch)
Group of OpenAI insiders warn of “reckless” race for dominance, calls for greater transparency and protections for whistle-blowers (The New York Times)

🤖 Machine Learning

🔬 Science and Breakthroughs

💼 Business and Marketing

Nvidia passes Apple to become second-most valuable public U.S. company by market cap (CNBC)
Core Scientific rejects $1 billion buyout offer from cloud computing firm CoreWeave (Coindesk)
Social app for creatives Cara grew from 40,000 to 650,000 users in a week because artists are fed up with Meta’s AI policies (TechCrunch)
Perplexity drops Hollywood-caliber fake movie trailer as its first ad (Adweek)

📺 Media and Entertainment

BNN Breaking passed off as a reliable news site, but it was an AI chop shop, publishing error-ridden content (The New York Times)
The state of gen AI in Hollywood (Variety)
AI music generators are being used to create hateful songs (TechCrunch)
Concerns rise around the use of AI in adult entertainment: bias, gender stereotypes, and more (BBC)

💰 Funding

Cohere reportedly has raised $450 million in funding from Nvidia, Salesforce Ventures, Cisco and PSP Investments (Reuters)
Seven AI nabs $36 million for its enterprise-centric AI cybersecurity platform (The Wall Street Journal)
Bem raises $3.7 million to automate unstructured data conversions for engineers (VentureBeat)
Storyblok raises $80 million to infuse its headless CMS with more AI (TechCrunch)

⚖️ Copyright and Regulatory Issues

U.S. Justice Department and Federal Trade Commission agree to divide responsibility for probing Nvidia, Microsoft and OpenAI (The New York Times)
Shutterstock’s AI-licensing business generated $104 million last year (Bloomberg)

💥 Disruption and Misinformation

Adobe responds to vocal uproar over Terms of Service language suggesting the company can “access, view or listen” to user content (VentureBeat)
Adobe removes AI images in the “style of Ansel Adams” after complaint (Petapixel)
AI anime flood: An infringement investigation of 90,000 images published to gen AI image sharing websites (Nikkei)
The World AI Creator Awards: The first AI beauty pageant (Wired)
AI is imitating the dead and dying, raising new questions about grieving (Associated Press)

🔎 Opinions and Research

OpenAI’s new research paper offers a look at the AI model powering ChatGPT (Wired)
Researchers find not all gen AI models don’t treat polarizing subject matters the same (TechCrunch)
How AI’s energy hunger upends IT’s procurement strategy (VentureBeat)

End Output

Thanks for reading. Be sure to subscribe so you don’t miss any future issues of this newsletter.

Did you miss any AI articles this week? Fret not; I’m curating the big stories in my Flipboard Magazine, “The AI Economy.”

Follow my Flipboard Magazine for all the latest AI news I curate for “The AI Economy” newsletter.

Connect with me on LinkedIn and check out my blog to read more insights and thoughts on business and technology.

Do you have a story you think would be a great fit for “The AI Economy”? Awesome! Shoot me a message – I’m all ears!

Until next time, stay curious!

Subscribe to “The AI Economy”

Exploring AI’s impact on business, work, society, and technology.

The Five-Stage Journey for Companies to Achieve AI Maturity