Enterprise hits and misses - the CES hype rollercoaster is robotic, and enterprises get a progress report on AI fluency

Lead story - Does your organization speak fluent AI? 

Sounds like a top question for 2026, eh? And it's the one Stuart poses - and revisits - in Do you speak fluent AI yet? How organizations need to get the corporate tongue around some people-centric challenges on the road to the Agentic Enterprise (Update - 9th January). 

Stuart has updates... He also recaps the questions on AI fluency he posed to Salesforce executives last year: 

The question I posed to them was this: HR has long protested its lack of a seat at the top table in terms of shaping business strategy. Agentic AI looks like a case of being careful what you wish for as it should free up resource to allow HR to do more outside of the mundane day job. But, I asked, is HR ready to meet and make the most of this opportunity?

With the release of Salesforce's AI Fluency playbook (free sign up required), it's time to revisit these questions - and the fresh takes are interesting. Stuart: 

To date, Salesforce has spoken mostly about its ‘four Rs’ approach to agentic adoption - re-design, re-deploy, re-skill and re-balance. Today it launches its AI Fluency Playbook, which according to the pitch blah blah is: 'A practical guide for businesses to prepare their workforce to confidently collaborate with AI to give employees agents and drive business impact at speed and scale.'

I'm all for practical guides; my biggest theme for enterprise AI in 2025 was the persistent gap between vendor AI visions and customer realities. I found that customers (mostly) buy in to where vendors are going. But as I said in my CRMKonvos hot seat video appearance last week, the nest question is: "What the heck do I do?"

Stuart explains Salesforce's take on AI fluency: 

The thesis behind the new framework is:

As companies look to become an Agentic Enterprise, success will depend on their workforce's ability to harness and apply agentic AI in their daily work. Businesses that build AI fluent workforces will drive greater growth and position themselves to attract top talent and become a best place to work.

Organizations need to build out AI fluency in three ways, argues Salesforce:

AI Engagement: Enhance employee sentiment and confidence around AI.

AI Activation: Drive consistent adoption in employees’ daily work.

AI Expertise: Build proficiency of human, agentic, and business skills that drive successful human+agent adoption.

I recommend Stuart's full take on the playbook, but here is one pull quote: 

All good and sound thinking, but clearly the potential is there for AI not to consume and destroy existing jobs, but to create new, more interesting ones for human beings. Equally clearly, there will be casualties along the way - not everyone will survive the transition to an agentic workforce. There will be those who are left at the roadside on this organizational journey. They will be the minority, perhaps, but that is inevitable.

Indeed. Salesforce's maturity model is welcome; we need more of these. But as I said in my What Enterprise AI Actually Wins At research reveal with Andreas Welsch, we need more: customers need to take a philosophical position on whether AI is for headcount reduction primarily, or if they are for augmenting/re-skilling employees, or where they land in between - and communicate that take clearly. 

Then, there is the need for precision on what's working in enterprise AI, and what's not. I run down my complete list on the show, but probabilistic LLM agents have a potent new set of pros and cons, and also require evaluation frameworks, to cite one need. As Alyx writes below: "Documentation quality, auditability, and governance determine whether systems can be deployed responsibly."

Net-net - if you want employee trust and AI buy-in, you're going to have to earn it, not mandate it.

And, as Stuart notes: you're also going to need new metrics, a topic Rebecca covers off well in Into 2026 - some AI realities that lie ahead as we correct some key agentic expectations. 

I think we'd best buckle up...

Diginomica picks - my top stories on diginomica this week

• Cyber-insurance underwriters are rewriting the rules – and identity security is the new price of entry - Alyx has an identity security wake-up call for us: "[Insurers are] not asking 'can someone break in?' anymore. They're asking 'how fast will you notice when they do?' That's why they care so much about continuous monitoring and anomaly detection rather than just perimeter defenses."
• What I’d say to me back then – be kind, not nice, says Epicor’s Kerrie Jordan - This latest entry in Madeline's standout series struck a chord: "Be kind, not nice means tell people what they need to hear because it's going to help them in the long run. You don't have to be nice to everyone, but you should be kind and that's not always easy." A pretty darn good mantra to live by in 2026...
• Your Questions Answered: AI, S/4HANA Implementation, and Cloud ALM - My fave podcast structure? Live questions from customers - without a net. In our latest podcast collaboration with ASUG Talks, Josh Greenbaum, Geoff Scott and I take on the top questions from ASUG Tech Connect submitted by attendees (that we didn't get to last time around). One curveball attendee question, via ASUG Talk's Jim Lichtenwalter, was simply: "ERP?" I thought we should answer that - does ERP as a category have a shelf life? 

Vendor analysis, diginomica style. Here's my top choices from our vendor coverage:

• Evergen and UiPath show how agentic AI tests process discipline in regulated industries - Alyx has an incisive use case from UiPath that frames enterprise AI in 2026 as well as I've seen: "Evergen’s experience offers a useful counter to the dominant agentic AI narrative, which often frames progress in terms of autonomy, speed, and scale. In regulated environments, those measures are secondary. Documentation quality, auditability, and governance determine whether systems can be deployed responsibly."
• Into 2026 - better AI means no more people silos! A finance and procurement view from SAP's Etosha Thurman - As I wrote my kickoff piece, I thought the neglected theme wasn't data or process silos, but people silos. Queue the enduring theme from the high/lows of 2025 events: "No, it's not AI - not exactly. Rather, it's how AI compels action from stagnation - action on data silos, process silos, and yes, people silos. I covered data and process in my 2025 AI project review. But the best events of 2025 provoked the people side of this conversation as well: better enterprise AI requires better collaboration."

CES blowout coverage - Is CES relevant to the enterprise? Most years, I find that CES is just an overdose of gadgets and gimmicks, with little enterprise appeal. But this year, there was a twist: I thought there was more enterprise overlap than usual, centered around robotics. Stuart had that angle covered, via: 

• CES 2026 - loving the machines at Caterpillar as AI turns steel into silicon and the physical digital
• CES 2026 – can Boston Dynamics’ Atlas robot carry the humanoid industry on its shoulder? 

But I thought it was Stuart's spicy CES kickoff that really set the tone for the CES techno-rollercoaster: 

It’s going to be a long week, isn’t it? 

Jon's grab bag - Stuart had a couple retail reviews for us: Plenty more e-commerce fish in the sea for Ocado apparently, but will it be served up with yet more 'jam tomorrow'?, and Primark cites doubling down on digital in Europe, but just can't shake off that bricks-and-mortar mindset. Oh, and if you didn't catch our diginomica network podcast, check Mark Chillingworth's customer 1:1 in The diginomica network podcast - Thomas Kleine-Möllhoff, CIO of Gardena.

Meanwhile, Chris is back on the AI/copyright policy beat in AI and copyright – technical proposals and legal absurdity abound in UK Government Inquiry: "Indeed, Google has essentially created a walled garden around the world’s content and is now renting it back to us, with rising numbers of Google searches now ‘zero click’ – i.e., the answers are consumed within AI search rather than the user clicking out to the external data source." I would add: and doing so with concerningly unreliable AI search results: ‘Dangerous and alarming’: Google removes some of its AI summaries after users’ health put at risk. 

Best of the enterprise web

My top seven

• 11 Runtime Attacks Driving CISOs to Deploy Inference Security Platforms in 2026 - Louis "zero trust" Columbus surges into 2026 with a series of strong posts: "CrowdStrike's 2025 data shows attackers breach AI systems in 51 seconds. Field CISOs reveal how inference security platforms defend against prompt injection, model extraction, and 9 other runtime attacks targeting enterprise AI deployments." Also see: Columbus' notable December post, OpenAI admits prompt injection is here to stay as enterprises lag on defenses.
• When “Safe Hands” Aren’t Enough: Why Implementation Risk Still Ends Careers - UpperEdge has your unvarnished project advice again: "The most successful programs I have seen do not rely on trust or control alone. They rely on informed, experienced challenge applied at the right moments."
• What can we learn from bad managers? - Vijay Vijayasankar hits on one of the hottest/thorniest non-AI topics of 2026: "From an employee’s perspective – what characteristic makes them think of their boss as a bad manager the most? I would think “micromanagement” wins that prize in a landslide." Bingo.
• Multi-media enterprise AI binge: I've got a slew of hot seat content coming out, including the video review: A candid 2025 Recap and bold 2026 Predictions, with Thomas Wieberneit and Ralf Korb, and my audio wrap with Andreas Welsch, What Enterprise AI Actually Wins At (Jon Reed).

Beyond Rabbit Ears on a Black & White TV - Lora Cecere excoriates supply chain AI vendors for falling short again: 

What I see in demos from tech providers aiming to drive value through agents is automation of today’s processes. A better forecasting engine. Probabilistic modeling. A deeper inventory optimization approach. The better use of streaming data for sensing. Yadda yadda. So, many experts are applying new technologies to improve current processes. Yawn. I don’t see anyone attempting to redefine S&OP in the same way Netflix redefined entertainment. I feel that we are caught in the AI stupid lane, fiddling with rabbit ears on the black-and-white TV as the world of technology evolves around us. 

Spot on, though I'd argue that some of the more conventional AI projects (e.g. better forecasting engine) is not a bad way to notch some value and start, well, iterating and learning things. But if you stop there, then I'm with Cecere. 

Whiffs

The 'whiffs' section is supposed to be acerbic, but ultimately light-hearted. However, X has been whiffing in a way that is impossible to make light of, nor to ignore. So, I would just say check this Ars Technica piece for X's latest ugh. On to the more fun stuff... Let's do some CES worst-in-show, shall we? 

Frogger, anyone? 

Wall Street Journal gave Anthropic's AI vendor agent a spin... 

Not clear why Anthropic wanted to duplicate these misadventures, but, the details of the 'whoops' are instructive.... 

See you next time.. If you find an #ensw piece that qualifies for hits and misses - in a good or bad way - let me know in the comments as Clive (almost) always does. Most Enterprise hits and misses articles are selected from my curated @jonerpnewsfeed.