growth forum, identity, and a room full of people who got it

i gave a talk today at growth forum nepal.

the topic was identity verification and what it actually means for SME access to financial services. i have been thinking about this problem for months in the context of building helios, but talking about it in front of people who are living the other side of the problem, business owners, lenders, people trying to build financial products for a market where identity infrastructure is inconsistent, made it land differently than it does in a technical context.

the actual problem

here is the version of the problem i tried to convey.

a small business owner in nepal who wants a loan has to prove two things: that they are who they say they are, and that the business they claim to run actually exists and has the characteristics they say it has. both of these sound simple. neither of them is.

the identity verification part breaks down at several points. some businesses operate with ID documents that are inconsistent across databases. some operators have names transliterated differently in different documents. some of the documents themselves have quality issues that existing OCR and verification systems handle poorly. a global KYC provider is not going to invest engineering time in solving devanagari OCR edge cases for a market this size. so the edge cases persist, and the system resolves them the only way available: by rejecting ambiguous cases or routing them to expensive manual review.

that rejection or delay is not neutral. for a small business owner trying to access credit, a verification system that takes three days to manually review a document is often functionally equivalent to a verification system that says no. the capital need does not wait.

the interesting thing about talking about this problem to a room full of people who work in and around this space is that most of them know this intuitively. they have felt it from their side. what the talk added, hopefully, was a frame for thinking about where the actual failure is occurring and what a properly designed system would look like.

the room

i met interesting people today.

that sounds obvious for a growth forum but i mean it more specifically. the conversations i had on the side of the event were about problems i had not been thinking about in quite the right way. a founder building financial products for a specific segment who described the exact verification failure mode i had been modeling, but from a completely different vantage point. a person working in lending who articulated the institutional trust problem in terms i am going to use in the next version of the helios pitch.

this is the thing that events like this are actually for, not the talks. the talks create a shared vocabulary. the conversations are where the actual information exchange happens.

i came away with three specific things i want to revisit in the helios design and one introduction that i think will be useful. that is a good return on a day.

a note on what i said

i was careful in the talk to separate two things that often get conflated.

the first is the technical problem of identity verification, which is largely solved in isolation. good models, well-deployed, can verify identity documents accurately and at scale. this part of the problem is tractable and we know how to address it.

the second is the institutional problem of making verified identity trustworthy enough that lending decisions can be staked on it. this part is not a technical problem. it is an accountability and reliability problem. it requires a track record. it requires clear handling of edge cases and failures. it requires the institution relying on your verification to understand, at some level, what your system does when it is uncertain.

most of the interesting work in this space is in the second category. building systems that communicate uncertainty correctly. that fail in ways that are recoverable. that earn trust not through marketing but through consistent performance under real conditions over time.

that is also, not coincidentally, the most interesting part of what we are building with helios.

coming back

i am back in bhaktapur tonight with a few notes and a clearer sense of what the next few weeks of helios work should look like.

there is something useful about leaving your usual context for a day. the problems look the same when you come back. your relationship to them is slightly different.

we have work to do.