The long and the short of IT - the week in digibytes

Putting Meta’s White Whale out of our misery

Ding, dong the great white whale is dead! Or it will be soon.  Meta is finally pulling the plug on the Metaverse, Mark Zuckerberg’s folly which managed to survive year on year, its budget intact, despite haemorrhaging money in the process. $80 billion worth of money, it seems. 

The end result? Some sub-par 1990s video game-style animation - that seemed to struggle with legs! - and general population that isn’t, as of 2026, wandering around in VR goggles plugged into the Zuckerberg-ian version of The Matrix. 

A salutary lesson for tech’s very own Captain Ahab? We’ll see. Costly lesson if it is; even more costly if it isn’t.

Digitizing WeightWatchers for the Ozempic age

One of the long term digital transformation strategies that diginomica has tracked over the years has been the ongoing attempts by WeightWatchers International to rebrand and reposition itself for the age of apps, smart watches, and Pelaton bikes. That’s been a journey with a lot of ups and downs, to say the least, and now we enter the Ozempic age and a whole new set of challenges. The problem is, despite its digitally-centric refocusing kicking off well before COVID, WeightWatchers in 2026 still appears to be talking about some pretty basic stuff, despite a claim from CEO Tara Comonte that:

WeightWatchers is beginning to feel different, look different, and sound different.

Trouble is, seen that and heard that before.

The big news now is that mobile experience has been re-launched as of January to rest on “a newly re-built foundational infrastructure and modern code base” it seems. Coming up - a new “AI body scanner, new personalized modes to support different phases of the weight loss journey, a proprietary weight health score and expanded coach-led virtual meetings among other ongoing innovations”.

Comonte claims:

Nowhere has our commitment to constant improvement been more evident than in the future digital road map we've laid out for the rest of this year.

diginomica will be watching and waiting.

Greggs, data science, and app loyalty - a tasty filling for some non-flakey strategic thinking

UK readers will appreciate the guilty pleasure of a lunchtime trip to Greggs, the baked goods centric fast food outlet that can be found on most British high streets. There’s nothing like a Greggs sausage roll to cheer you up. (For non-UK readers, a sausage roll is pork sausage meat - or vegan equivalent - wrapped in a flakey pastry case, and is definitely not the same thing as a Breakfast Sausage Croissant before any of our US audience write in!).

What’s flakey pastry got to do with enterprise tech (apart from the flakey angle)?  Well, quite lot behind the scenes where, according to Executive Director Richard Hutton, there are “some interesting experiments going on with automated monitoring”. The firm is also in the throes of upgrading to SAP S/4HANA.  He adds:

In our support teams, technology is starting to help us with automation on desktops, new systems for customer and shop support, which are making the whole process more productive. It's meaning our teams can cope with the growth in the business without adding more resource. And increasingly, AI tools will support that even further going forward.

The firm has also been hiring data scientists to its ranks. They have been looking at the Greggs app data and performing more detailed technical analysis on the role of the app. Hutton says:

It’s become, I think, more of an essential as part of your mix, to have something which rewards the customers who are loyal to you. If it was driving the incrementally the data scientists are saying, then we'd be sort of shooting off the charts, wouldn't we, and the fact that we're not, I suppose, shows just how important it is in terms of securing the loyalty of your existing customer base. Whether it attracts new customers, that's always the heart of it rather than holding on to the customers you have, but I think it's a super important part of our armory.

Can the AI-enabled sausage roll be far behind?

Why AI models aren’t adding up for accountant job threat

You know all those panic stories about how AI is coming to take over everyone’s job functions, from lawyers thru marketers to accountants? Some useful data from vendor Dual Entry which released the results of a large-scale benchmark evaluating how modern AI models perform across real accounting workflows. The benchmark tested 19 leading AI models on 101 domain-specific accounting tasks, covering transaction classification, journal entry creation, bank reconciliation, financial reporting, and month-end close operations.

OpenAI GPT-5.4 achieved the highest accuracy at 77.3%. The second-best model, Gemini 3.1 Pro, scored 66%, more than 11 percentage points behind GPT-5.4. But most models scored below 65% accuracy across accounting workflows, while older models, such as GPT-4, scored only 19.8% on the same task set. Even the best performing model still fails roughly 1 in 4 accounting tasks.

Despite rapid improvements in reasoning models, the results highlight ongoing reliability gaps in financial automation: no model exceeded 80% accuracy, and most systems failed more than one-third of accounting tasks. This all leads to a timely - and salutary - conclusion from Dual Entry:

Large Language Models are increasingly capable at generating structured text, categorizing transactions, and drafting journal entries. These capabilities can accelerate repetitive accounting tasks such as first-pass transaction classification and draft financial reporting. However, accounting systems do not run on drafts.

But according to Financial Intelligence specialist MindBridge, CFOs are increasingly looking at implementing AI to stop financial damage, even though undetected errors and poor data quality are quietly eroding business profitability and creating a massive drag on operations. Nearly 90% of organizations report “critical delays due to data errors”, despite the push for AI to fix the problem, with 62% of businesses reporting a moderate-to-severe financial stemming from poor data quality.

He said what?!?

That was my Darth Vader imitation - "Impressive." Yes, I'm Jedi.

                                                                                        Jensen Huang, NVIDIA CEO.  

(Now, Jedi pretensions I can live with, but I’d have said Yoda as first choice, no?)

Product byte - Salesforce orchestrates with NVIDIA

At NVIDIA’s GTC conference, Salesforce unveiled its Agent Toolkit to bring AI agents directly into the flow of work via Slack, Agentforce, and NVIDIA Nemotron models. As per the announcement, the expanded partnership includes:

• NVIDIA Nemotron 3 Nano: Now available in Agentforce, this model’s 1-million token context window and architecture enables agents to reason across long customer histories and multi-step workflows at a fraction of the traditional compute cost. 

• The Slack “command center”: While Slack provides the collaboration and engagement layer, Slackbot acts as the coordination layer - receiving user requests to trigger Agentforce workflows, reason over Data 360 context, invoke Nemotron-powered processing, and orchestrate agent actions across enterprise systems. 

• Architecture collaboration: Salesforce and NVIDIA are also collaborating on clearly defining what happens in various layers to give companies a practical blueprint for scaling agents with Slack as the collaboration and engagement layer;  Slackbot as the co-ordination layer[ Agentforce as the reasoning, execution, and agency layer; Data 360 as the context layer; and NVIDIA AI infrastructure and Nemotron models as the accelerated AI processing layer 

He said what?!?

I have so much gratitude to people who wrote extremely complex software character-by-character. It already feels difficult to remember how much effort it really took.

                                    OpenAI founder and CEO Sam Altman. 

(Whether he remembered other people’s efforts as his models gobbled up other people’s work over the years is unsaid. But I could hazard a guess...)