Lead story – Earning AI buyer trust – where do we go from here?
My last piece sparked some discussion on the dilemmas of building AI trust. (What are the components of buyer trust for AI services? Time for an enterprise rethink!) My goal? Advance the discussion on buyer trust by compiling a list of trust requirements that would sufficiently challenge vendors – and customers.
The ensuing debate provoked the question: is AI trust more a matter of technical acumen or good old fashioned (but hard to find) buyer empathy? In the process, I did a rethink of my own view. A quick reminder: buyer trust in AI services is not the same as trust in AI itself.
When vendors fret “how do we get our customers to trust our AI,” they are obsessing about the wrong thing. You can get a very good result out of AI without fully trusting it. Arguably, you’ll get a better result, via the judicious use of guardrails and approval steps.
As it turns out, the components of AI buyer trust span both technical and business outcomes. For this quick rundown, I’ll pick two. Start with transparency:
Transparency in AI pricing, architecture, and data privacy/movement – is there a bigger factor in AI trust than visibility in how your data is being handled? Transparent pricing is a close second. Most vendor AI pricing is a gooey (and perpetually shifting mix) of embedded features, add-on credits, consumption pricing shifts, and outcome flirtations.
And: mix in explainability and observability. Lukas Die Kunst had a pointed response on X:
The part about vendors hiding their monitoring tools says a lot. In aerospace I instrument everything precisely because failure is expected somewhere. Admitting your AI needs observation should be a selling point, not an embarrassment.
— Lukas Die Kunst (@LukasForTech) September 16, 2026
After the piece came out, I realized we’re back to the paradox of AI leadership. Earning AI trust is about both domain expertise and technical chops. It’s about being able to demonstrate business outcomes, and minimize customer risk. It’s all of those things. If the AI provider is weak on any of them, then trust will be elusive.
Buyer empathy matters also – did I miss that? Before my last installment, yes. Even empathy – a decidedly human quality – can benefit from a tech infusion. Is there a better way to achieve empathy than by deeply understanding your customer’s data, and showing them how your services can help them – not generically, but tied into their proprietary data assets?
This buyer trust review will evolve – let’s hear from you on what else should factor in next.
Diginomica picks – my top stories on diginomica this week
Vendor analysis, diginomica style. Here’s my three top choices from our vendor coverage:
diginomica does Dreamforce (again) – our wall-to-wall Dreamforce 2026 coverage is there for your review. Here, I will pick just a few highlights from our own the ground team:
My diginomica pick for quote of the show: no question, it had to be about Salesforce’s aggressive interface strategy, which is perhaps best described as an interface-agnostic strategy, despite the Claudeforce partnership and the engagement around Slack as a human/team/agentic interface. Via Stuart’s wrap piece:
At Dreamforce, Salesforce showed off three live interfaces for its offerings, but that’s only the start, says Benioff:
‘By the time we get back here [to Dreamforce 2027], or let’s say by even as we get to January 1 of next year, I’m confident you’ll have a lot of different interfaces. The interface customers will choose their religion here. We’re agnostic. One big advantage we’re going to have is the agnostic aspect of the interface. We’re going to give customer choice. I think customers will have different religious preference here. There’s no question.’
A few more vendor picks, without the quotables:
Jon’s grab bag – Ian manages to tie the fall of Troy into AI data mischief in The fall of Troy and why enterprise AI buyers should beware geeks bearing gifts. Chris weighs in on the doomsday media overdose in Something for the weekend – why we should worry more about AI’s makers than the evil robots bent on human extinction. He also takes on the energy consumption debate in AI energy usage – is the industry scaling up when it should be scaling down?
AI alignment needs human aligntment; George has thoughts in Never mind AI alignment, what about human alignment? How XPRIZE is trying to incentivize this goal. Stuart points out that Microsoft issued an under-the-radar document, a humanist code of conduct on AI – a nice contrast with doomsday hyperbole. See: Cracking the code – why people do matter more than AI!
Finally, Alyx is back from an instructive deep dive at AGNTCon – MCPCon Europe:
Soria Parra said that when MCP was being created two years ago, tool calling “barely worked. Models had to be constantly corrected.” The protocol and platform layers have come a long way since, but the maintainers building them say the skills you need to work with them were already there. [What the people building MCP say enterprise leaders are getting wrong about AI skills].
Best of the enterprise web
![]()
My top six
AI agents exploit ungoverned credentials – perhaps the idea that AI agents thrive on ungoverned credentials is obvious, but too many organizations have left more than a crack in that door open. Louis Columbus is on the case.
3 Critical Questions to Ask Your SaaS Vendor – UpperEdge’s Adam Mansfield has three key questions to press with SaaS vendors. I’d say we’re off to a strong start with: “Can You Show Me the Value I’ve Been Getting from the Products I Already Pay For?”
Tuesday’s Tips: Five Common Sense AI Regulations – Constellation’s Ray Wang wants to get us unstuck from regulations paralysis.
Business Value and the Limits of ROI – Esteban Kolsky put a big twist on the ROI argument: “ROI is therefore a poor primary measure of transformational business value when the objective extends beyond improving an existing activity. Legitimate returns from automation, lower costs, or faster processes show that an activity improved, but they do not establish whether the enterprise created a better operating model, changed how decisions are made, developed new sources of growth, or built capabilities competitors will find difficult to replicate.”
Your AI agent failed. The model might not be the problem – To get further into enterprise domains, agentic failure needs to be pinpointed – and traced: “‘It’s not enough to just look at the logs or the inputs and the outputs,’ Hallak tells The New Stack. ‘It is important to figure out how it got to the answer. What were the reasoning traces? What tools did it utilize? Where did it get stuck? Where did it decide to try a new approach?”
The Neuro-Inclusive AI Playbook (Christina Fischer) – Andreas Welsch had put out one of the better AI podcasts of the year.
Whiffs
This isn’t going totally awesome:
So Microsoft wants AI agents… to play video games for you, so you can… get back to your AI First workplace?
Microsoft Vibe Gaming: an AI to play your games for you pivot-to-ai.com/2026/09/09/m…
“What if gaming, but we remove the bit where you play the game?”
-> makes sense to me, have to get ready for agents to live our lives entirely while we watch from comfy couches…
— Jon Reed (@jon.diginomica.com) 2026-09-13T23:17:28.980Z
Singapore is testing out a plan… to pay humans to be more literate:
Hmmm… 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.