Get ready for agentic AI - Rethinking the Hype Cycle #3
The robots are coming, whispering through modems and riding cockroaches
Hello👋
Welcome to Rethinking the Hype Cycle, your practical guide to AI and what's next in tech for people-first leaders.
If you've joined us from Responsible AI Leaders newsletter, welcome! This is a free article. Future issues will be for paid subscribers. Sign up now for new posts and member offers.
Thanks, Simon for this rave review! A free subscription coming your way.
These last several months have got me deep thinking about how my tech consultancy intersects with people's lives and our businesses' interests. Right now, the technology we buy is from big tech firms, many asserting unreasonable control over nation-states. Some believe the tech broligarchy wants to shift America to a quasi-corporate-style management by removing those pesky obstacles like people and democratic decision-making that stand in the way. 🏛️
Even if you write this off as another conspiracy theory, as tech buyers, we need to think carefully about whether Meta, Microsoft or Google are businesses we want to do business with. I've taken X (and Musk adjacent) off this decision list, as the ethical vs value review is open and shut.
Tech journalism is a way we can decipher the extent of destruction. It's not tech or politics. Tech IS now politics. As legendary tech journo Mike Masnick of Techdirt writes, "This isn't about politics — it's about the systematic dismantling of the very infrastructure that made American innovation possible."
In this newsletter, I will sometimes cover stories where technology and politics overlay to help you make informed decisions on what you do and who you buy from. And yes, I still Google and use Gmail. Weaning off Meta slowly.
Rethinking the Hype Cycle is here to help you survive the hype. Grab a cup o' chai (or something stronger) and remember: if you're not working on the bleeding edge of technologies, you don't need to bleed.
On to the trends. 👉
AI and frontier tech trends
Agentic AI has arrived
Agentic AI is "like putting your brain in a jar." And it's made of thin glass.
That's what privacy-focused messaging app Signal's president Meredith Whittaker thinks about this new paradigm of computing, where AI does tasks for us — without prompting.
Whether you agree or not, agentic AI is the trend you'll hear as much about as generative AI (GenAI) this year. If you don't follow Zoe Scaman, get to it #1. She wakes us to the opportunities and threats of new tech and how they impact women. Here, she heralds the potential and perils of the oncoming agentic AI era.
I’ve recently been talking tech with school and college students. I tried to explain to the younger kids that technology isn't just Metal Mickey (or today’s equivalent) tin robots. (My robot cat, though, was a big hit😺) This is a nice explainer on how robot agents evolved from "softbots" to today's AI operators. Are we at the start of a 10-year cycle from hardware to intuitive and integrated software?
Agentic gets a boost from Manus. It's described as a general agent that can answer more of your queries in one chat. Registered in Singapore, but will its Chinese origins limit them? It's not quite a Deepseek moment, but the rise of good enough AI from the east is getting US big tech twitchy.
Agents need convincing voice modes
A convincing voice you want to interact with will be key to getting people comfortable with using agent models. Enter Sesame, the new ultra realistic voice agent channelling serious Scarlet Johannsen 'She' vibes. Not sure how much I trust a product that describes itself as an uncanny valley but it does sounds good.
I was convinced by voice agent Boardy, described as the best-connected AI. After a 5-min interview about what I'm about and my challenges with a calming Australian voice, it feed back a succinct view of what I'm about (actually about not what I do) to the point I felt heard. Must not think AI hears me, but it did understand and fixed me up a connection. A few quibbles: no surnames or LinkedIn shared. I can see this tech taking off to give professionals advice, connections and more.
This Jibberlink demo is super cute. Two AI agents talk to each other to get the task done. Their agent-to-agent “voice” sounds like a ‘90s modem dial-up. You’ve got mail.💋
AI Agents that can navigate the awkward
If you're not following Henry Coutinho-Mason
get on it #2. I avoid big events in the main as I find it almost random if I get to meet anyone that I can connect with. Henry talks about getting an AI assistant to help you navigate big conferences. Find people, summarise what you missed, and overlay who you are talking to so you don't goof it up. Great way to use AI to get value from IRL networking.Faking the crocodile tears🐊
MJ Crockett, cognitive scientist at Princeton University, on why many tests claiming AI is more empathetic than experts are bogus. Consider this when you think about replacing your customer-facing folks with AI. And let's dispel the jibber coming from tech experts, including the increasingly dubious Stephen Fry, that AI could have a conscious. That’s science fiction, not fact.
Will AI be America's financial ruin?
This economist predicts that the AI hype will hollow out the US economy and jobs, making the cost of starting a business so astronomical that only big enterprises will make gains. Surely that's not their grand plan?🤔
The writing's on the wall for technical scribes
AI is producing technical docs for the pharma industry. Novo Nordisk claims Claude is now reliable enough to slash teams of 50 technical writers to three. Heavy emphasis: this is not a good time to get into a career in business writing. ✍️
AI (and AI slop) is all around us
Gemini AI and search are melding. It will power Apple search and there are now experiment with an entire takeover of Google search. AI slop is now inescapable in your search engine. However, if you want results sans AI overviews, try adding "-ai" to your search query. You're welcome. 🔍
We should be concerned about the impact of AI slop on audiences: So much online is untrue, fake or lazily produced, which makes gaining trust that much harder for brands who want to do intelligent content properly. LA Times used AI to write a completely fake article about the KKK. These media credibility misfires will make audiences mistrusting, as well as interference of tech broligarch’s in editorial as is the sad fate of the Washington Post and it’s ‘editor at large’ Jeff Bezos.
Each day, another platform kicks the AI slop bucket. Pinterest is swimming in it, "It's just like finding needles in haystacks now for real art" 🎨
To tame the turds, I reviewed my Facebook Content preference settings. I could select how much low-quality clickbait, unoriginal, problematic and sensitive (including violent) content I wanted to see. You can turn it down, but not off. And wildly, you can turn it up! Facebook's main purpose is now a data-harvesting platform. If you pay for targeted ads, it’s throwing your coin in the slop bucket.
AI subs as valuable as Netflix
I think of LLMs as business tools. But numbers show that consumer subs are on the rise and are as sticky as your Netflix subscription. Based on Netflix's current output, it has the potential to be more entertaining. As many people have heard of AI as Doritos and Viagra. 🍆
Altman eyes up World app domination
Altman's World startup is coming for your eyes. Like Musk attempted with X, Altman's tech empire has a vision of creating an 'everything' app. If you go to China, you'll experience this with Alipay where cash rarely exists. Altman's twist: building a payments app based on iris-scanning to prove you're human. Broligarch and your eyeballs. What could possibly go wrong? 👁️
Please, Mister, lend me a dime to invest in AI
🎻Get your tiny violins out for AI investors: it's hard to make cash from AI when investors want predictable revenue, and it's all entirely unpredictable. It's a similar part of the hype cycle we've seen for technologies before, like the commercial web. We're likely to see a fair number of AI firms and system providers crash and burn in the process.
Every day I'm hustling (for AGI)
Sergey Brin checks notes 5th richest man in world running checks notes 3rd biggest tech firm in world says they're so far behind with creating AGI that checks notes will do engineers out of a sustainable career so engineers should hustle a 60-hr work week. It's like starving to give away all the crops before a famine. Tech used to be cool, with nerds who wanted to give it their all to make great stuff not feeling compelled to. When did Silicon Valley go all Wolf of Wall Street? 🐺
Smart shapeshifting robots
Robots that can shape shift from fluid to solid structures. The ultimate Transformers.
I'm mad for alternative and scrappy robot concepts. So ingenious: Why build a tiny robot for disaster recovery when you can cellotape a camera to a cockroach? 🪳
It sure beats this faceless musculoskeletal robot with an advanced range of motion. Not at all creepy. 😳 It's not entirely clear what it’s for, but it concerns me that these robots emulating humans could be used in nefarious ways. Warfare tech budgets grow by the day.
Who is leading the design skills for AI?
More experienced designers and freelancers are leading the path for AI adoption. And on their own time. With little leadership from their agency. Potentially, those with the most to lose from the march of AI in design. And they should be concerned. Fiverr Go (as hustled by the Del Boy of tech, Steven Bartlett) launched. It lets designers ‘clone’ their style for others to use for a micro fee. Could we see the automation of all low to midweight creative briefs?
Tech regulation, data security and brand safety
Most prefer discussions in moderation
A global survey shows the majority are in favour of moderation and removing abusing content rather than its abandonment under the 'free speech' banner. Sadly, many expect to get abused when participating in social media. Americans are most reluctant to consider removing abusive content.
UK: Unions up the ante with protecting creative workers from AI
UK creative industries union demand protections against "rapacious tech bosses". My new favourite alternative name for the AI broligarchy. Celebs glam up the copyright abolitionist debate as stars released an album of studio reverb sound without the actual music. And it's still Paul McCartney's best album in years, IMHO. 🎵
And this analysis shows parking AI data centres in the UK isn't benefiting jobs, business or the environment. There is limited evidence that AI is working any magic to benefit public services yet. Glad my spidey senses are on form - I've said before, the value for money vs. job creation didn't stack up from my state-funding days.
US: Wild wild tech
Ditching responsible AI, mass firings at NIST (National Institute of Standards and Technology) – America's new kings are deregulating tech, crypto and ditching measures to keep us safe. Welcome to the wild, wild west. 🤠
Digital land has reverbed with the shock that DOGE, the (allegedly) government efficiency chainsaw squad overseen by Musk, has culled 18F. For those familiar with the pioneering work the UK government did in making government websites accessible and affordable by bringing in the talent in-house, this is the US equivalent. This clearly won't save the state anything as contractor agencies will do less for substantially more.
Quantum tech up the security protection
Quantum is heating up. Now a Swiss chip claims it can offer protection from quantum cyberattacks. 🔐
Regulators: Listen to the kids
Those regulating and making decisions on AI in their business: listen to the kids. They want cleaner, transparent and ethical AI for their futures. See The Children's AI Manifesto from Turing Institute. 👧
From my talks with kids, they're hyper-aware that AI is coming into their lives, and they’re feelings ain't all good. In the US, Mechanical skills are back in fashion for kids as the AI threat for white collar skills looms. Kids are more disengaged than ever, reports The Atlantic, as they see AI as supplanting the need to learn about Pythagoras or algebra. I remember this as an 80s kid about the calculator. But AI is too tempting as a 'homework machine’.
If 92% of UK students are using it, surely we need to educate them in the rights and wrongs. I believe AI is a great learning aid if the temptation to give you all the answers without trying can somehow be unpicked (or fairly tested for).
Careful does it for AI adoption
So much good juice in this short and zesty read from the formidable tech for good team at Careful Industries on AI adoption and avoiding the overwhelm: "It's possible to be both careful and curious about technology adoption". 🍊
A load of Grok
No need to read (it's paywalled), but I'm furious about this FT headline. It implies that Grok, which pulls on the (non-quality or safety controlled) X output for uncensored recommendations is for brave brands. Businesses should be thinking damn hard about risk, disinformation and safeguarding or fall foul of various laws and protections, not least the EU AI Act. Who's paying your editorial bill, FT? And in a surprise to no one, Grok, the so-called “free speech” platform, blocks anti-Musk/Trump sentiment. 😠
A Lush digital detox
British ethical beauty brand Lush walked outta the big tech social platforms to build their own. Here's how it's going. It's a shame that few other brands had the bravery to follow suit. Building your own is a trend for further investment.
Investing in your platforms and mailing lists is ultimately the most sustainable form of marketing. I love this call to arms from a child of Geocities to build your website free from corporate shackles. Started my career as a Dreamweaver gal. 🕷
Legal AI: You be the judge
AI may be good enough for routine legal admin but it lacks reasoning insight for proper legal work. Will deep reasoning and RAG (retrieval-augmentation generation) to sync lookups to your in-house information change that? ⚖️
when I hit the dashboard
Next year, the EU will require cars to have buttons & dials - not just touchscreens - to get a top safety rating. A small victory for design usability. How about we bring back smarter phones with buttons that older folks and others can use without getting help? 🚗
AI business use cases
If coding isn't your vibe, try vibecoding
The best explanation and demo yet on vibecoding and finding the right inflection between autonomous (computer does it all) and AI-enhanced (I tell computer what to do) workflow.
Golden AI-arches
Maccie D's uses predictive maintenance AI and smart kitchens to destress their joints. I can't imagine how stressful it is working in a hyperspeed fast-food restaurant. Dismiss them at your peril: McDonald's is as much a tech and innovation firm as it is a place to get your Big Mac fix. 🍔
Moving beyond automation and productivity to capability
Impressive capabilities with new AI models can create new code, games, and scenarios. Professor Ethan Mollick says that instead of asking "What tasks can we automate?" leaders should ask "What new capabilities can we unlock?" 🔓
BCG, shout to their C-suite homies:
"Most companies aim too low, prioritizing smaller-scale, productivity-focused initiatives. Leading companies AI investments are reshaping key functions and inventing new offerings."
I've been reading about the pros of embracing freer rather than structured thinking. Salmon Labs riff on the joys of AI for embracing messy thinking and data for strategists. And to stop AI from rotting your brain, keep asking, Is AI helping you think quicker, or more deeply?
And we need to tune up our little grey cells. Research from Microsoft no less and Carnegie Mellon shows that confidence in AI outputs numbs our critical thinking skills.
How to train your AI for intelligent automation
I wrote for the dev community at Stack Overflow on what you need to know about using APIs for training AI models, what's termed "intelligent automation".
AI shouldn't steal the joy
You've probably seen the meme from sci-fi writer Joanna Maciejewska that AI is writing poems and songs but leaving us with chores like doing the laundry.
When technology takes away the work people enjoy, don't be surprised if they disengage or even go into full “Neo Luddite” machine-wrecking mode. Here's another example from Google LLM trying to do the wrong part of the job.
"For many scientists, generating hypotheses is the most fun part of the job. Why would I want to outsource my fun to a computer and then be left with only the hard work to do myself?"
Let's talk about ChatGPT, let's talk about you.com and me
In AI model news:
A button-down summary of what each ChatGPT tier offers. And free is now, for many use cases, pretty good. Security, less so. And with a decent non-profit programme.
Big may ultimately be better in the long game for a dominant model. ChatGPT is leading the wolf pack. 🐺 Despite or even because of the interest in Deepseek, active users swarmed by 1/3 in 2 months.
And OpenAI is cleaning up comparisons for deep research. 1 in 3 said the results of the reasoning models matched the results of an experienced hire. Is the end nigh for early-career researchers?
Note: This isn't an endorsement for OpenAI; I've been using Deepseek, Claude, Notebook, You.com (a model that draws on the best of other models; seek out a free trial) and others for different use cases. Here's a handy chatbot analysis to work out which one is best for your use case.
Want to tame the firehose about OpenAI news with expert analysis? Check out What Did Open AI Do This Week? From TBD Network, a great biz and tech network I'm a member of.
I heartily recommend the TBD Annual Conference on 3 April. Eigengrau, making sense of the state of grey we're in. Online. Pay what you want. No excuses. I have 1 free pass for a Rethinking the Hype Cycle subscriber. Reply to your subscription email if you're interested and can attend on the day, No replays.
How to hire for AI fluency
Great idea for hiring managers: timed tests in AI fluency help see if candidates understand how to tease results from AI. Meanwhile, HBR on AI and your job search: Your job will be eradicated by AI, and no human will be reading your CV. So don't use AI to apply, or it'll be flagged as AI slop. Is this the enshitification of hiring?
Uber for hired guns
"Our bodyguards got us matcha."
Uber-style gig worker service launched for armed security on demand with influencer videos. This may not be as niche as you think, it's perfectly timed for chaos times. Despite the theme, gig platforms to access skilled service can be a good thing. Or not.
I’ve talked in prior editions about the gig platform hiring nursing and pushing salaries down. Brace yourself. Now, Mercor, a $2Bn valued tech startup's 21-yr-old founder wants to carve up full-time jobs into gig work pieces. If you’re buying from these services, think about the merits and ethics. Could you develop your own AI agent to hire people direct at more equitable rates and cut out the broker?
➡️What to do next to get ahead
Get ready for agentic AI by mapping your AI value chain
If you’re already overwhelmed getting your people on board with AI, agentic AI will blow the roof. It’s time to plan and bring some buckets.
🪣 1 What can be automated, streamlined or simplified?
🪣 2 What can be augmented, improved and optimised?
🪣 3 What can be semi-automated – a human touch on an automated process?
Then you’ll have more answers about what tools you need for each challenge and where agents may sit in your mix.
Create your own (wo)man-machine model
The big wins won’t be from taking the fun and collaborative bits out of people’s jobs and handing it to a smooth-talking agent (while they quietly die in an open-plan office cubicle 2-hour journey across town). Design workflows that blend the best of both. Look beyond productivity metrics to more meaningful outcomes. If your team feels excited or curious rather than threatened by AI, you're getting the balance right.
Explore alternatives to big tech
The US broligarchs aren't the only players. Consider European alternatives that may offer better alignment with EU values and regulations. There are alternatives to the Google / Microsoft monopoly.
Explore European digital services and cloud alternatives. Investigate open-source models that give you more control and transparency. For specific use cases, narrow AI providers often outperform the giants. Start by identifying one system you could migrate as a pilot. Benefits extend beyond avoiding monopoly risk – you may find providers responsive to your specific needs.
In case you missed it: Previously on Rethinking the Hype Cycle
Women are swimming in the slow lane with AI adoption
I updated my AI adoption research for International Women’s Day. Good news: Lots of new research into gender and AI, and intersectionality. Bad news: The further you are from being one of the Silicon Valley tech bros designing AI, the less likely you are to reap its benefits.
What to do now and next with the EU AI Act⚖️
Is AI Literacy the best move right now? If you've put your head in the sand about this compliance, I've got you.
Bonus 🎊
Some fun things to remind you tech isn’t all mind-numbing productivity or job-erasing dystopias.
Remembering the good ‘ol social media days
I’ve recently presented my wiggly arts-grad to tech consultancy career at a high school (6th form in UK money). Got all nostalgic finding my first Dreamweaver sites on the Wayback Machine. And this song You’re No One If You’re Not On Twitter from Ben Walker aka I Hate Mornings. Ben celebrates being on the Twitter tribe in the days we freely shared before rage-fuelled algorithms and spammy ads ruined it all.
Ben played at an unconference I organised in Nottingham where people from the business and cultural communities shared how social can connect business, communities and each other. *Shaking fist at sky* damn you, tech broligarchs! The internet was good, and there are pockets of good and greatness still. Follow me on Bluesky to spark those early-Twitter-days discovery vibes.
Love you, bae! Saturday Night Live does AI learning. My ears cannot process any more of those Notebook "podcasts". If the tech has made it to a SNL parody, the shark has jumped. 🦈
Your regular reminder: How to avoid tech hype
Hype rewards itself without social returns or meaningful technological progress. More wise words again from Signal’s Meridith Whittaker, a voice of sanity in the responsible tech space. And we’ve closed the loop.
Until next time,
Susi O'Neill
EVA trust in tech
www.evadigitaltrust.com
People-centred tech and AI adoption 🤝 Transparent communications 🎙️ Tech talks and inspiration. Get in touch.