Investor James Cham urged documenting early integration strategies.
We don’t honestly know the best approaches to rebuilding companies around AI agents, especially in ways that expand competitive advantage & augment existing human capabilities. Practical agents are merely months old. Experimentation (and productive failures) will be required.
Positive users highlight productivity gains and solo builders achieving more with AI agents through experimentation, while negative users point to high costs, unreliability in complex flows, and questionable competitive advantages.
@ChachaMarquis Meh. There are obviously ways to increase competitiveness from higher productivity, higher intelligence, automation, augmentation, speed….
@ChachaMarquis Most companies have positive ROI already
Historically the experiments and failures of trialling such advanced tech would be limited to a very technical audience.
By the time I’d have got hold of it (in sales and marketing), it would normally be far better conceived and established. However technical I thought of myself, my role just didn’t afford me the time and skills to be anything other than a consumer of tech.
But now a new set of users like me has been exposed, trialling and failing with things that change so frequently in a way that hadn’t been possible before. Taking on novel solutions. Being imaginative with problems. Crowdsourcing ideas and giving it a go. Not just waiting for a feature in a product.
It’s a learning curve not to over engineer a solution at the expense of everything else. I see experienced devs able to shrug off the excesses an agent might encourage them to pursue and therefore get a lot more done.
But it’s better to try and fail than to have never tried at all. It develops skills I can see lacking amongst GTM teams and their leadership (who haven’t yet been thrust into AI at the same velocity as engineering).
You can now build rather than passively consume solutions to problems. And that’s a really fascinating culture and expectation shift.
@emollick we don’t know that there are any competitive advantages at all. there likely aren’t. the assumptions around AI economics have always been a little deranged
Yeah I think a lot of the work is tied to identifying for a company what the infinite backlog and forelog looks like. Like if you could easily move at 5x the speed or have 2-3x the capacity what would you meaningfully apply it to? I think that question is what causes a lot of people to pause and hesitate. It’s solvable but takes some intentionality and sit down thinking that I think most people feel/think they are too busy to engage in. Too busy to plan, too busy to save time, sort of thing.
@emollick My favorite company built around agents is @every! They write a lot about their approach https://every.to/p/after-automation cc @danshipper @kplikethebird
@emollick @Afinetheorem Yes
@emollick And cost efficiency. Long-form agentic flows are way too expensive now.
(and no, this is not an AI response lol)
@emollick but not obvious enough for anyone to have found any yet particularly relative to the costs of use. got it.
@ChachaMarquis @emollick yeah, definitely no advantages to faster, autonomous, smarter processes doing things behind the scenes.
The best way to think of onboarding an AI agents is exactly like when onboarding a new human employee. You have to make sure:
1) you give them ramp time till they hit prod (Think Sandbox). 2) you give them resources for both functional and cultural understanding (think memory and context) 3) you give them a buddy to help navigate any bottlenecks (Think tools/connectors) And 4) you only graduate once they meet a basic threshhold after the end of ramp time (think evals)
Humans employees never worked on prod systems from Day 1, Ai agents also should'nt.
Hope this helps!
@emollick A parallel would be when the media was swamped with computers for sale. Everybody needed one but it seems games was the only real reason for families to own one.
@jazzplane @emollick not if everyone has the same thing on hand. and seeing how marginal the benefits have so far and how high the costs are i’m not sure why we are still assuming we are halfway to Blade Runner.
@ChachaMarquis @jazzplane @emollick Every factory got a steam engine (and later electric generators) and they made a difference during those early industrial revolutions.
Those companies all benefitted even though everyone had the same thing.
@emollick think that’s quite dated and clearly not accurate giving how recently the labs started actually charging firms for the true cost of AI. total opposite at every firm i’ve spoken too including our own where they’ve committed 1b to this for very little real world benefit.
@emollick .@KSimback I wonder how many in Mollick's circle know about OpenServ's work in this area?
@lorepunk @emollick @every @kplikethebird ❤️❤️
@emollick Not to mention completely unreliable after a certain number of chained steps!
@emollick True. But we can take a really educated stability at it given that the automation play in general is pretty comprehensively fleshed out.
@emollick You’d be surprised how far enterprises are going with agentic workloads. It’s pretty extreme; faster than any adoption rate I’ve seen in my career.
