Enterprise hits and misses - the AI circus faces off against user trust, transfo...

the AI circus faces off against user trust, transformation meets reality, and speculations on Google and HubSpot clog social media Enterprise hits and misses - the AI circus faces off against user trust, transformation meets reality, and speculations on Google and HubSpot clog social media

Lead story - The AI circus versus trust and harsh realities

The potential and problems of gen AI are tough to unravel from each other. But in the enterprise, trust and data quality are paramount to adoption - and results.

Rebecca gets the ball rolling with some eye opening stats in: Roll up, roll up! The AI circus is open for business with trust center stage in the ring:

In Valoir’s recent report, Language Matters – AI User Perceptions, we found that the AI trust battle is just getting started.

In our recent study, we found that 84% of workers say they have experimented with generative AI, either on their own or at work. However, although there has been plenty of press about the current and expected benefits for users of AI, potential users are skeptical. In fact, 17% of workers believe AI is about as useful as a screen door on a submarine at work, and only 15% thought that AI could help them jumpstart a writing task.

Some of these dismal percentages are likely the result of using early versions of these tools; now part of the AI trust problem is luring users back for another go. The trust issues don't stop there:

Privacy violations top the list of worries, with 51% of workers expressing fears about potential privacy violations by AI systems. Additionally, apprehensions about AI acting autonomously without human intervention (45%) and the perceived threat of AI replacing human roles (38%) add layers of complexity to the trust equation.

Valoir found that "trust" is also tied to AI brand credibility, but that's a complicated picture too, with some vendors (e.g. Google and Microsoft) winding up on both lists (trust and don't trust). So how do we get to a different place? Rebecca concludes:

So, in the immortal words of PeeWee Herman, what is all of this supposed to mean? It means that trust will be key for both buyers and potential users of AI, the battle for trust is just beginning, and it’s anyone’s to win – or lose.

In the enterprise, much of this will play out on a per-industry basis, as Stuart documented in Retailers excited about gen AI; retailers don't have the data foundations in place for gen AI - Salesforce study exposes some harsh realities. Though retailers are optimistic about use cases like personalization, pesky data issues cloud the AI picture:

That’s the positive. The negative bring us back to that damn data issue. Drilling down from the headline conclusions, the underlying detail doesn’t get any more comforting. The study finds that only 17% of respondents reckon to have a complete, single view of their customers. Nearly half (49%) are still in the preliminary stages of building or even considering the creation of a complete customer data profile.

AI trust is best examined by use case - with the degree of human oversight a factor as well. It is easier to imagine trustworthy AI for personalization than for autonomously approving mortgages. Medical use cases are more appealing when AI is used to assist in diagnostics, not for determining care eligibility.  Not trusting AI for the latter feels like more like a virtue than an error. As for retailers, Stuart notes the upside as well:

Personalization and customer service use cases are cited as target areas for generative AI adoption by retailers. The study finds that just over a third of retail employees are using gen AI today, with that expected to rise to 45% by the end of next year. Of that percentage today, some 93% say they are already using generative AI around personalization, in the form of email copy or product recommendations to customers.

Personalization can certainly drive sales; no arguing with the data there. But as I hashed out with Thomas Wieberneit last Friday, AI-based hyperpersonalization is no simple thing to achieve.

Diginomica picks - my top stories on diginomica this week

• Revolut Bank CIO invests in change - Mark Chillingworth profiles a CIO with a transformation agenda: "Product-based teams that include technologists working with product owners, UX designers and business line operations are becoming popular, but as Peñacoba says, in traditional banks, a new way of working and new types of teams are hard to embed with people who have worked a certain way for 30 years. So, what is the role of the CIO?"
• Why Amex GBT embraces AI with executive oversight - George examines an aggressive AI play, albeit with plenty of adult supervision: "Amex GBT’s example of appointing AI leads across engineering, finance, customer, and employee sides is a smart move. It also has a greater chance of balancing the incredible amount of AI hype against practical opportunities to generate business value safely."
• HR, meet AI - everything changes...perhaps - HR is one of the most promising - but also problematic - areas for HR. Cath digs in: The upshot is that HR headcount is unlikely to fall much. Instead, AI will likely be used to automate transactional activities, leading to the creation of 'augmented roles.'"

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

• Google Cloud President - ‘We can be the adult in the room when it comes to AI’ - With Google Cloud Next on deck, Derek delves into Google's AI stance: "Renner says that customers coming to Google Cloud are asking for help with prioritization when it comes to their AI projects. They are seeking help with trying to understand where success lies and how they can develop a strategy." Watch for Derek's updates on the ground from Google Cloud Next this week...
• Domopalooza 2024 - how customers are using Domo dashboards, Bricks and AI - Alyx shares customer proof points from Domo's user conference: "Domo Bricks can be configured and tailored by front-end developers to match branded styles, and are being used by WKS to deliver insights across around 370 stores, on the move."
• Meeting the needs of Marketing and Digital Experience - a conversation with Contentstack CMO Gurdeep Dhillon - Barb provides an update on Contentstack's own strategy - can Contentstack succeed in the emerging "composable DXP" space?

Oracle event coverage and use cases - post CloudWorld London, Oracle stories kept coming: 'Build once, use many times' - UK Crown Representative Nick Griffin on how firms like Oracle can help improve government service delivery (Stuart), and, fresh back from an Oracle NetSuite event in NYC, Brian filed SuiteConnect NYC - helping customers scale with Oracle NetSuite.

A few more vendor picks, without the quotables:

Jon's grab bag - I have not heard of the Milk Shed buyer predicament. From how Martin describes it, I'd prefer not to get milked anytime soon: The 'Milk Shed' model - how enterprise buyers can avoid being semi-skimmed.

The long tail of established users committed to an old technology and content to be `milked’ for years is fading fast: the pace of change in IT means the long tails are getting rapidly docked.

Indeed. Finally, I took another crack at Avasant Research's gen AI project report in "Your generative AI is your new brand" - so get it right! Lessons from Avasant Research's report on real world gen AI: I asked the authors: "How do you think gen AI costs will impact adoption? Will vendors effectively bundle these costs into existing licenses or affordable add-ons? Or will cost be a factor that limits implementation of gen AI solutions until cost/operating expenses are reduced?" Their answers were revealing - though gen AI costs are going to be a moving target for a while...

Best of the enterprise web

My top seven

• Defending against IoT ransomware attacks in a zero-trust world - Ransomware attacks are getting more sophisticated, but Louis Columbus is on the case. Can zero trust tactics work with IoT? In a word, yes: "Enforce least privilege access for every endpoint and IoT device."
• This year's top 8 use cases for AI, and what tech professionals need to support them - Joe McKendrick parses data on the gen AI use cases in play for 2024 - and the skills needed to make the happen. I was glad to see the older-school "predictive maintenance" on the list. My two least favorites are content generation, where the line between useful and mediocre/overconfident/buzzword-infected is razor thin, and so-called intelligent search, which can be good IF it is well architected (e.g. tied to a knowledge graph), but can also run the risk of generating inaccurate results if utmost care isn't taken (see: Bingbot fails).
• Why ERP Projects Fail: Finding the Gaps in Your Program Plans - Over on UpperEdge, Ted Rogers issues some salty/field-tested questions customers should press vendors on.
• Predicting Digital Transformation Failure: 5 Telltale Signs - Eric Kimberling has a few things to say on failure as well, this time on digital transformations gone awry: "When vendors wield disproportionate influence, projects become vendor-centric rather than client-centric."
• The OEMs’ Dilemma with Software-Defined Vehicles -  Hybrid vehicles have hit a few speed bumps of late. Evangelos Simoudis breaks down the issues: "Customers, especially non-Tesla owners, started complaining about their BEV experience due to inadequate charging infrastructure, higher vehicle repair costs, and changes in their vehicles’ residual value due to price wars among automakers."
• Ten years later, Facebook’s Oculus acquisition hasn’t changed the world as expected - Yeah, and one look at that cumbersome headset explains why... But Meta isn't done playing the long Metaverse game, which to be honest makes things more interesting, especially if you are a shareholder.
• As deal rumors fly, Alphabet and HubSpot would be a strange pairing - I debated putting this in hits/misses at all, given this is much more about speculation than anything tangible yet. But, Ron Miller did a good job of framing why Alphabet could be interested and the obstacles ahead. But as Thomas Wieberneit told me in this video highlight, the end of the third party cookie (and the impact of first party customer data) is a big storyline here.

Whiffs

On the more serious side of whiffery:

But we could fill this column with that type of thing most weeks. I can't say the same for this pitch, which Brian Sommer thoughtfully forwarded to me:

(via Brian Sommer)

Thanks for throwing my hat in the ring, Brian! I don't think I can top that one, so let's quit while I'm behind...

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.