Anthropic's AI research: automation vs augmentation

Ultimately it is always about the human and machine working together.

Here's a question for you, Jeffrey Bussgang: Does success require that augmentation extend beyond "good-enough" situations to times where the results need to be near-perfect? I see why startups are willing to take the risk of AI. What other choices do they have in a time-constrained, personnel-constrained moment in their development? What about other parts of the business lifecycle where 80% right isn't good enough?

Any time I’m doing some monotonous or time consuming task, I now often find myself thinking back to LTV and you pushing all of us to become AI native. So far, I definitely see myself in the “augmentation” bucket - I’m able to work on much more simultaneously than I ever thought I could

Other than the jobs that are fully listed task based like finish a-z and the job day is over, augmentation will almost always win over automation/AI. AI will help boost the efficiency and speed where humans will use AI vs. the cases where AI alone or AI assigning tasks to human as it simply do not know enough randomness which is must for innovation and asking new questions what we do not know yet. All AI systems are trained on what we already know and the knowledge is out there on the internet so it may not be courageous or wild enough to think totally unthinkable as we have done for centuries to come up with new ideas, products and developments in science.

Thanks for sharing, Jeffrey

You should always level up your skills. We at Zero Wall Street (https://zerowallstreet.com) are using Anthropic models for coding, and with every new release, it comes with challenges, making it harder to work with. You should be at least one level creative if you want to keep your position, grow, or develop new products.

I think there's definitely a messaging problem, for lack of a better term, with this. It's worth noting that the 2025 State of Tech Talent by the Linux Foundation also tells a different tale than the narrative dominated by automation, but that narrative doesn't come in a vacuum. Besides media reports about jobs being lost to this, the big names/personalities in this space lead with sound bytes about how "AI is coming for (fill in job)" or about how it can do the work of X number of (fill in job). Or you have Andy Jassy flat-out telling people headcount is going to go down because of AI. All of that carries weight in public sentiment. So with that, no question augmentation is a story not being told enough, especially to those outside the tech space who have more intuition into this, and I would add that it's probably an uphill battle to tell that story with not enough effort being made to do it at the moment.

Diana Rocha

Thanks for this, Jeffrey Bussgang. You'll be interested in this Rebekah Bastian

Absolutely. The most transformative AI impact we’re seeing isn’t in replacing talent but in amplifying it. The leaders who design for augmentation and not just automation will scale faster and smarter.