This is a story about Was Not Was, or AI Not AI. Every week or so, Paul Greenberg and Brent Leary conduct what they call a Happy Hour with a group of CRM analysts and enterprise types. Somehow we got into a discussion about the history of the new journalism, or new media. The Happy Hour is not recorded, the theory being that the players can let down their hair and be more open about the things we are going through in this era of work from home not home and other stuff that frankly should be talked more about not less. But I am in the minority of that view so I mostly listen and very occasionally start talking about Trump until I get deported to the kid’s table. A good proxy for talking tech but not really is Musk and his team, which is now interestingly merging with the constant rise of AI Media. The Signal crisis of the week neatly rolls the moment up. I’ve been getting schooled on the players by Gangster member Keith Teare, who has been diving deep in his role as a member of the venture community.
Mostly I’ve been pretending that AI is fake news. The truth is I’ve been waiting for a there being there other than the rich get bigger and we’re all be happy. As I sit and listen, I see the outline of something blundering up the hill. It’s partially about the notion that you can use these tools to write code for you. A little AI intern sitting there just waiting to assist you in fighting off the bad guys and finding a place in the knowledge not knowledge economy.The other stuff (generative AI) I snarl at, but the idea you can use freeware code to restructure your life into meaningful building on top of conversations with imaginary safe associates and avoid layoffs is almost enough to not give up completely and wait out the next Reacher.
Let’s not ignore the land grab that’s going on. The social hairballs that have metasticized into cash machines inevitably suggest there will be a reckoning where all this free data will get tagged with open not open in the rush to synthesize training data. DeepSeek was the whites of their eyes moment (don’t worry, Perplexity and co. will soon offer starter kits where you trade data for sea shells.) No need to recognize the names; just work on how to ask the questions.
Let’s say you want to figure out how to get more efficient with the cable (now AI) news networks. YouTube TV has this 4way grid screen of MSNBC, CNN, Fox, and BBC. It’s unlikely Google will watch how many times I lose interest in the repetitive and partisan to no end framing on the right to center spectrum, but of course they have the data. If they provided a viewer’s choice channel that rolled up a leaderboard of favorites, those analytics would quickly rise in the charts. But will it happen anytime soon? Look at Substack for the view from the bleachers. The startup, funded in large part by Andreesen Horowitz and the new Mar-a-Lago kitchen cabinet, seems to be content to stick with culture and the deprecation from advertising of their subscription model. Anytime you think they might be encouraged to see what happens if they are under pressure to move to a more hybrid model like subscriptions plus sponsorships, just look at the old media they supposedly replace. Graze the lower thirds and you see the partisan veneer driving the parties’ agenda and not the customers’ desperation at surviving the extreme positioning. The only reason I doubt this move toward progress is that incumbents think fear works better. The only reason rationality might happen anyway is competition from China and maybe Europe.
None of this landed remotely near reality on the happy hour. But, sick of talking AI, Brent and Paul invited a few of us to talk about anything other. Here are the questions we should have been asking AI:
Since we can’t talk about AI (Trump) what can we? AI seems not to like prescriptive questions, the ones I think of like tell me exactly what questions to ask and then give the answers to an intern to make it so. I quickly discoovered that AI will tell you what steps to take but not provide a mechanism for typing the data into the form until it works. It apparently will accept followups that lead to debugging by reading the error messages and commentary back into the script. Better would be it does that talk for you without being asked. Also hire (provide auditing by a group of people and pulling out group questions and suggestions that proceed in parallel.) Upsell this data to YouTube to incentize the creation and inventivising for that particular group of customers and provide bundling of their favorite sites as a Costco-like subscription perk. This is akin to the old trade press where subs were freewared to decision makers and resold to vendors for funding and then to Gartner etc. for analysis.
Why will this not happen? Publishers are the biggest blocker, as they view the impact of the economy as separate from the analysis by the users. This is a subtle directive to engage the digital worker Benioff meme. Not that he doesn’t believe it, but where is the moat to protect Salesforce and SAAS in general?
If the AI media is driving, what does it look like? Develop tools to tease out path to questions to build analytics around survival, work from home, leaderboards of successful question building. connectors to bridge answers to form entry (agents). What are the blockers to agents? Are they real? What are the real collaboration blockers? Build series of questions about WFH home blockers, team dynamics and UI, competitive secrecy, political, nation state dynamics? Here they come. Walking down the street. Hey hey we’re the Monkees. A complete unknown.
Share this post