Lead story –When AI can build anything, these are the things you should ship – and how
I’m fixing to blow an agentic gasket. Why? Because I think the automation storyline is overdone. Yeah, I said it:
- I don’t think it’s going to be as easy to automate deterministic systemswith probabilistic technology as so many of my peers do, however much “context” we attempt to apply.
2. Routine work should have already been automated by (more affordable) rules-based automation. Much of the “routine” work that is left to automate is not as routine as AI evangelists think, even with “agentic reasoning” to help… It’s going to require some human (and system) supervision, especially when multiple agent–to-agent handoffs come into play.
3. Agentic automation requires monster work in data quality, human collaboration and use case design. It is absolutely valid and useful, but: what if our autonomous talking points obscure the real action?
- Building cool sh%t (and nifty skills/agents).
- Creating agentic “advisors” that can answer contextual workplace questions all day long, and execute needed tasks out of that context. In our eagerness to take every process under the sun away from human hands, are we overlooking the power of text-based UIs, giving a whole generation a permanent respite from tedious screen-by-screen navigation?
Mark Samuel’s latest piece pertains to building cool things. He kicks off:
When almost any idea can become a feature, deciding what deserves to reach customers becomes a significant challenge and a potential risk. At a panel session at the recent Harness Unscripted event in London, engineering leaders discussed how organizations can embrace experimentation and ship effective products.
Yep – experimentation is key. But if we don’t ship, we don’t eat:
The panel recognized that AI makes software easier to build; the hard part is ensuring humans of suitable standing remain very much in the loop. As James Ford, Principal Engineer at price comparison specialist Compare the Market, said, AI increases the speed of development, but quality remains a human consideration:
‘The scarce resource is human attention and proper judgment over things. We can build stuff faster, but the scarcity is getting human eyes on this 10x that we’re doing.’
Failing to maintain tight control over the order of releases will mean organizations struggle to exploit potential increases in coding speed:
Fair enough – so what’s the lesson learned? Stop cranking out 10x code with tech debt risk? Build smarter?
The panelists did not downplay the benefits of experimentation – and they cited examples of how business users, including project managers, can surface the best development ideas with agentic tools.
But from there, we’re back to degrees of automation, not a blanket, one-size-fits-all approach to autonomy. The dreaded operational drag of humans-in-loops doesn’t have to be a drag. Bringing humans in at just the right time, – either for internal and external-facing process – can help you strike a better balance of experience and efficiency than leaving your stakeholders to put up with processes they can’t influence.
AI might make it easier for developers and non-IT staff to push products live, but that doesn’t mean they should. Everyone who touches the software development lifecycle must recognize that prototyping is the first stage in a longer process – and the human in the loop is key to ensuring outputs are safe and secure.
That’s a better starting point than build-fast-and-break-things, or, the dreaded thing so many brands do now: build-fast-and-ship-substandard things. Why not start a bit tighter, and then loosen controls where it makes sense?
Diginomica picks – my top stories on diginomica this week
- AI slop – an inevitable result of lazy Marketers or something that creatives can challenge? – Stuart bears down on the grittier side of AI in Marketing – is AI slop a bug, or a feature? Right now, it sure seems like both…
- Tokenomics – curating new cost and trust dimensions of enterprise AI – Chris continues our tokenomics series, via the slicing and dicing of several new reports.
- As agentic AI evolves beyond its visionary phase, is enterprise adoption with the pragmatists now? – Barb riffs on new research from G2. It’s not just about out-of-the-box agents, but building your own: “Organizations are looking for platforms that have a pre-built ecosystem of connectors, actions, or skills, as well as a marketplace.”
- Why it’s time for a software conference course correction – as the road miles pile up, Brian’s getting a tad grouchy about the enterprise event circuit: “We got earfuls of AI technology infrastructure discussions and almost nothing about business value.”
Vendor analysis, diginomica style. Here’s my three top choices from our vendor coverage:
- A brazen move to take a troublesome rival out of the game? What’s behind Salesforce’s $5.7 billion gambit to buy Braze? – Stuart breaks down a Salesforce shopping spree of some magnitude. Impact? “If Magnusson’s thesis is correct and his expectations around Braze’s pipeline are realistic, then removing this player from the board does look like sound game play on Salesforce’s part.”
- Re-invention all over again as SAP pushes past €10 billion revenue in a record quarter- Stuart on a big proving ground quarter for SAP, with the SaaSpocalypse hyperbole looking more foolish by the day. Deep customer footrpints (and deep datasets) turn out to come in handy with AI. And yet, there is plenty of work ahead – a good quarter is just that.
- ServiceNow World Forum London – how Experian and BT are taking different paths to agentic AI- fresh use cases from Madeline: “From knowledge graphs and audit logs to people-centered AI strategies” – sounds like my kind of AI project stories…
The diginomica roadshow rolls on – once again we fanned out, in search of customer validation and whatever insights we could find amidst the keynote bombast
IFS Unleashed – Phil and Brian were on the ground in Orlando:
- IFS Unleashed – AI agent workflows at Kodiak cut out friction and speed resolution – Phil
- IFS Unleashed – data as the success key, the opportunity area, and unspoken friction – Brian
While Kodiak only has a few agents deployed today, Buhigas sees the number rapidly escalating, perhaps even into hundreds by next year, with some of the auditing and checking of their output being automated by agents. Inevitably that will ultimately lead to changes in the work that employees currently do.
Several more show reviews of note:
- Waves 2026 – Make sets out its ‘glass-box’ vision for deeper enterprise AI adoption – Ian
- NVIDIA GTC Berlin – humanoid robots continue their march into the automotive sector- Chris
- Wrike Collaborate 2026 – AI raises the stakes for Wrike’s Work Intelligence Graph- Ian
A few more vendor picks, without the quotables:
- Why Unit4 bought Napta – helping PS firms plan resource and skills in times of change – Phil
- Why do UK professional services firms post the highest margins of five surveyed markets and trail the US on retention, according to new Certinia research? – Alyx
- Confluent tells buyers to use AI budgets to fix data estates – and backs open Streamhouse category – Derek
Jon’s grab bag –Stuart has the news we shouldn’t have missed in The long and the short of IT – the week in digibytes. He also has a thing or two to say about the state of AI in the EU. For starters, check: Why is Brussels seemingly so hellbent on making sure that EU-Inc can’t compete with the US?
Mark Chillingworth looks at how enterprise leaders become founders in What CIOs and CTOs learn when they become founders and investors. Sarah looks at the intersection of iGaming and AI in Something for Weekend – fancy an AI-powered flutter? Ethically, of course!Finally, Katy presses the right question in What if AI just perpetuates gender inequality in manufacturing? (“Clearly, this technology is not neutral. Data reflects cultural norms and areas with a lack of data also reflect cultural norms” – indeed).
Best of the enterprise web
- AI Is Changing Everything About Hacking. No One Knows What Happens Next – I’ve been insistent: with modern systems and zero trust security practices, AI agents are not the Doomsday harbingers they appear. But there’s no question that many crucial operating systems rely on legacy components, and are therefore nowhere-near ready for the relentlessness of the agentic loop attacks. As per 404 Media: “So many people discovering that the whole Internet is dry tinder and the only thing keeping it from burning is the fact that teenagers haven’t found the matches yet.” Yikesy. Also see: Paul Kedrowski’s The Hackening Has Begun, where he argues that most of today’s interconnected, “poorly maintained” software is a soft target.
- Gartner Survey Finds Only 13% of CIOs Say Their Enterprise Has Realized Significant or Transformational Value from AI– this story made waves this week, though it’s really just the latest in a fairly long line of similar studies. We’d be remiss if we didn’t spend more time studying the smaller percentage of companies that are achieving real value. The (still kinda rare) successes are arguably more instructive than the underperformers.
- Open models gain traction in enterprises, tech CEOs say – this one from Constellation’s Larry Dignan seems like a no-brainer, but what it means for the increasingly frantic
frontier model companies is the interesting bit. - Reflection and Celebration for the Supply Chains to Admire Methodology– let’s step away from the LinkedIn circular bowl
maelstrom of AI for some supply chain nuggets from the ever-readable Lora Cecere. - Why No AI ROI? How to Overcome the Three Pillars of the Productivity Paradox – an old school skewering by Joshua Greenbaum, albeit with the new school topic of agentic ROI firmly in scope.Josh Greenbaum “Here’s the productivity growth problem in this study: Though the AI scribe made physicians more efficient in moving patients through the patient queue, with no attempt to measure accuracy or the improvement in patient outcomes there’s no way to truly ascertain whether there was any business ROI, much less a medical ROI...”
Multi-media pick – CRM Konvos co-host Thomas Wieberneit boiled my latest hot seat appearance on their show in The Agentic AI Trust Gap: You Own the Results, Your Vendor Sells the Tokens. (You can see the full CRM Konvos replay on LinkedIn, which had some spicy audience questions). Wieberneit writes:
Jon’s list from the road put two items at the top of customers minds: pricing and ROI, then interoperability and user experience. How do Copilot, SAP and ServiceNow agents work together? Vendors lead with the next agent instead. That discrepancy is not an accident.
No, it is not. On we go…
Whiffs
I just love practical solutions to vexing real world problems:
It seems like Meta almost gets offended when they get left out of the whiffs section… But give them credit, they come back strong:
Oh, and I’ll have six of the…. Hmm, this is awkward:
See you next time… If you find an#enswpiece that qualifies for hits and misses – in a good or bad way – let me know in the comments asClive(almost) always does.Most Enterprise hits and misses articles are selected from my curated@jonerpnewsfeed.
