Flair BuchiMiracle Okonkwo

Work01

Xpend mark

AI agents / financial infrastructure / crypto payments

Xpend

What happens when software can actually move money? Xpend is how we're answering that: crypto that can be spent like ordinary money in Nigeria, with agents that help, and controls that keep people in charge.

Role
Technical founder / CTO at JustXpend AI.
Status
Active product. Still being built.

What it is

Xpend helps people deposit digital assets and spend them as ordinary money in Nigeria: bank transfers, bills, airtime, invoices, and related everyday payments. It shows up in places people already use, including WhatsApp and a mobile app.

On the surface it looks simple. Underneath it is a harder question than chat design.

The interesting problem

Crypto makes value programmable. AI agents make software able to understand requests and take action. Put those together and you get a new kind of product, and a new kind of risk.

An agent that can prepare a transfer, find a recipient, or start a deposit is useful. An agent that can move money without clear limits is dangerous.

So the question that shaped the work is:

  • How much financial autonomy should software have?

How we approached it

We treat understanding a request as different from being allowed to move money.

The agent can clarify, gather details, and prepare what should happen. Spending still requires an explicit confirmation from the user. Convenience lives in the conversation. Authority lives in the confirmation.

At the same time, the product has to hide crypto complexity. People should not need to think about wallets, chains, and settlement mechanics just to pay someone or buy airtime. That complexity still has to be handled correctly behind the scenes, or the money is wrong.

  1. 01

    Request

    Someone asks for a financial action

  2. 02

    Understanding

    The agent interprets and may ask questions

  3. 03

    Prepared action

    A clear proposal of what will happen

  4. 04

    User confirmation

    Explicit approval before funds move

  5. 05

    Execution

    The payment runs only after that gate

  6. 06

    Outcome

    Success, failure, or cancel, reported plainly

The parts that matter

A few ideas carry more weight than a feature list.

  • Control before autonomy

    Agents become more capable as they can do more. Every new capability has to respect the same boundary: prepare freely, move money only with permission.

  • Confirmation before trust

    Money is not considered available just because a payment looks like it started. The system has to know when value is actually settled enough to spend.

  • Providers as infrastructure

    Banks and payment providers fail in different ways. That is not a detail each feature should reinvent. It is an infrastructure concern: how the product survives when a rail is down or slow.

  • Simple for people, exact for money

    Users should feel like they are using money. Operators still need the details underneath to be precise. Those two needs pull in opposite directions.

Related work

There is a merchant side of the same problem: businesses that want the utility of digital assets without becoming crypto infrastructure companies. That work lives in Xpend B2B.

Full write-up: /work/xpend-b2b

What I work on

As technical founder and CTO, I help shape the product and the systems underneath it: how agents are allowed to act, how deposits become spendable balance, how payments reach real rails, and how the consumer and merchant sides relate.

Serious systems are collaborative. I do not claim every line. I do own the responsibility for the design direction and the hard trade-offs.

Where it is now

Xpend is active work at JustXpend AI. The product is real, and the difficult parts are still difficult.

I keep coming back to the same question: how intelligent software should be allowed to touch money when mistakes are expensive.