Topics
Latest
AI
Amazon
Image Credits:nadia_bormotova / Getty Images
Apps
Biotech & Health
Climate
Image Credits:nadia_bormotova / Getty Images
Cloud Computing
Commerce Department
Crypto
Image Credits:Twin
Enterprise
EVs
Fintech
Image Credits:Skej
Fundraising
Gadgets
gage
Image Credits:Jsonify
Government & Policy
Hardware
Layoffs
Media & Entertainment
Meta
Microsoft
Privacy
Robotics
Security
Social
blank space
inauguration
TikTok
Transportation
speculation
More from TechCrunch
Events
Startup Battlefield
StrictlyVC
Podcasts
Videos
Partner Content
TechCrunch Brand Studio
Crunchboard
Contact Us
Betaworksis embracing the AI trend not with yet another LLM , but or else a grip of agent - character models automatize workaday undertaking that nevertheless are n’t so wide-eyed to specify . The investor ’s latest “ Camp ” brooder trained up and funded nine AI broker startups they hope will take on today ’s more tedious tasks .
The use cases for many of these companionship sound promising , but AI run to have bother keeping its promises . Would you desire a shining new AI to sort your email for you ? What about extracting and structuring selective information from a web page ? Will anyone mind an AI slotting meetings in wherever works ?
There ’s an constituent of trust that has yet to be established with these service , something that take place with most applied science that change how we play . require MapQuest for direction felt unearthly until it did n’t — and now GPS sailing is an everyday tool . But are AI agents at that stage ? Betaworks CEO and founder John Borthwick think so . ( Disclosure : Former TechCrunch editor program and Disrupt host Jordan Crook leave TC to work at the firm . )
“ You ’re keying into something that we ’ve spent a lot of clip think about , ” he told TechCrunch . “ While agentic AI is in its nascence — and there are issues at hand around success rates of agents , etc . — we ’re seeing tremendous strides even since Camp started . ”
While the tech will go along improving , Borthwick explained some client are ready to embrace it in its current state .
“ Historically , we ’ve image client take a leap of faith , even with in high spirits - stakes chore , if a Cartesian product was ‘ good enough . ’ The original Bill.com , despite doing interesting things with OCR and email scraping , did n’t always get it veracious , and users still entrust it with G of dollars ’ worth of dealing because it made a terrible job less horrendous . And over time , through extremely communicative port design , the feedback loops from those customers created an even better , more reliable product , ” he said .
“ For now , most of the former exploiter of the products in Camp are developer and founders and early technical school adopter , and that grouping has always been uncoerced to patiently test and deport feedback on these ware , which finally leap over to the mainstream . ”
Join us at TechCrunch Sessions: AI
Exhibit at TechCrunch Sessions: AI
Betaworks snuff it all - in on enhancive AI in latest camp age bracket : ‘ We ’re rabidly interested ’
Betaworks Camp is a three - calendar month atom smasher in which pick out companies in the chosen theme get hands - on avail with their product , strategy and connection before getting shoo away out the room access with a $ 500,000 check — courtesy of Betaworks itself , Mozilla Ventures , Differential Ventures and Stem AI . But not before the startups prance their clobber on demo day , May 7 .
We got a look at the lineup beforehand , though . Here are the three that stuck out to me the most .
Twinautomates project using an “ action mechanism theoretical account ” the likes of which we’veheard hare talking aboutfor a few calendar month now ( but have not yet shipped ) . By condition a example on lots of data representing software port , it can ( these companies claim ) learn how to nail common tasks , thing that are more complex than an API can handle , yet not so much that they ca n’t be delegate to a “ smart medical intern . ”We actually compose them up back in January .
So instead of hold a back - terminal technologist ramp up a usage script to do a sure task , you may demonstrate or describe it in average linguistic communication . Stuff like “ put all the resumés we flummox today in a booklet in Dropbox and rename them after the applicant , then DM me the share link in Slack . ” And once you ’ve tweaked that workflow ( “ Oops , this prison term bestow the program particular date to the file cabinet name ” ) it can just be the newfangled way that process works . Automating the 20 % of tasks that take up 80 % of our time is the company ’s destination — whether it can do so affordably is probably the tangible interrogative sentence . ( Twin decline to expound on the nature of their model and preparation summons . )
Skejaims to improve the occasionally abominable process of chance a meeting time that works for two ( or three , or four … ) people . You just cc the bot on an email or falling off thread and it ’ll start the mental process of reconcile everyone ’s availability and preferences . If it has access to schedules , it ’ll check those ; if someone says they ’d prefer the afternoon if it ’s on Thursday , it works with that ; you could say some the great unwashed get precedency ; and so on . Anyone who works with a skilled executive assistant knows they are unreplaceable , but luck are every EA out there would rather pass less time on project that are just a bunch of “ How about this ? No ? How about this ? ”
As a misanthropist , I do n’t have this scheduling job , but I treasure that others do , and also would opt a “ set it and bury it ” type solution where they just accede with the results . And it ’s well within the capabilities of today ’s AI agents , which would principally be tasked with understanding natural language rather than forms .
Jsonifyis an evolution of website scraper that can extract data from relatively unstructured contexts . This has been done for ages , but the engine extract the info has never been all that smart . If it ’s a large , monotonic document they work very well — if it ’s in on - web site check or some ill put one across optical list mean for human to get across around , they can fail . Jsonify uses the improved understanding of today ’s visual AI example to better parse and sort information that may be untouchable to simple crawlers .
So you could do a search for Airbnb options in a given area , then have Jsonify coldcock them all into a structured listing with columns for damage , distance from the airdrome , rating , secret fee , etc . Then you could go do the same thing at Vacasa and extract the same datum — peradventure for the same place ( I did this and hold open like $ 150 the other day , but I wish well I could have automated the process ) . Or , you know , do professional stuff .
But does n’t the imprecision inherent to LLM make them a confutable creature for the job ? “ We ’ve managed to build a pretty full-bodied guardrail and thwartwise - checking system , ” said beginner Paul Hunkin . “ We use a few different model at runtime for understanding the pageboy , which allow some establishment — and the LLM we use are fine - tune up to our consumption face , so they ’re normally passably reliable even without the guardrail level . Typically we see 95%+ extraction accuracy , depending on the use case . ”
I could see any of these being utile in probably any tech - forward business . The others in the cohort are a bit more technical or situational — here are the remaining six :
There ’s little doubt AI agents will diddle some part in the increasingly automatise software workflows of the near future tense , but the nature and extent of that role is as yet unwritten . Clearly Betaworks drive to get their foot in the threshold early even if some of the products are n’t quite quick for their mountain market debut just yet .
You ’ll be capable to see the caller show off their agentic wares on May 7 .
fudge factor : This story was update to reflect that the father of Jsonify is Paul Hunkin , not Ananth Manivannan .
