Hivemind Labs · Australia

People before AI

Hivemind helps Australian businesses make better use of their people's time and capability. We understand the work first, find where technology belongs, and stay responsible for designing, building and managing the solution we recommend.

Ben and Jake, the Hivemind founders
BenJake

An AI partner for the way your business actually works.

We work with businesses that see experience, judgement and relationships as valuable. They don't need someone to sell them an AI tool. They need a partner who can understand how their business works, recognise where technology can genuinely help, and stay involved in making sure it works.

We find the work technology should be doing.

Every business has work that needs to happen but doesn't need the full capability of the person doing it. Whether that's chasing information, moving data between systems, routine correspondence, reconciliation, scheduling, triage, status checks or follow-ups.

We find that work, understand why it exists, and determine the right way to improve it. Sometimes that means AI. Sometimes automation, integration or a simpler process is the better answer. Where technology belongs, we design, build and manage solutions around the way your business actually operates.

Your best people should spend more time on work that needs them.

When experienced people spend their time on work that requires little of their experience, the business loses more than hours. It loses capacity for better decisions, stronger relationships and the higher-value work their expertise makes possible.

People before AI means putting technology to work where it creates genuine value, while focusing people's time and capability where they matter most. The goal isn't to automate as much as possible. It's to give people more time for the work that needs them, and help the business achieve more with the capability it already has.

  • Judgement, relationships and the final say stay with your people.

  • Technology earns trust by being reliable and open to scrutiny, not by being clever.

  • The goal is for your people to become more capable, not more dependent.

We don't start with AI.
We start with your business.

Before recommending a platform, model or automation, we understand how your business works, where people's time is going, what is getting in their way and where the strongest opportunities exist. Then we decide what should remain human and where technology can make a meaningful difference.

Only then do we determine the right solution. We design it around your processes, integrate it into the way people already work, and define clearly where technology can act and where people remain in control.

And we stay involved. We monitor, manage and improve what we build as your business, systems and needs change, so it continues to deliver value over time.

  1. 30 minutes

    An insight call.

    With Jake or Ben. A straight answer on whether it is worth going further.

  2. 5 days · fixed price

    A Working Week.

    A workshop with the people who do the work on Monday. A working prototype on your own process by Friday, with a quote in writing.

  3. Ongoing

    A partnership.

    Built into the systems you already use. Your people keep sign-off. We stay responsible for what we built.

How we work

We do the research, not just the work.

Alongside client work we publish our own research and the practices we have learned running agentic AI systems: how they should remember, check themselves and stay safe to leave running.

  • Agentic memory: what an agent should keep, recall and let go of.

  • Safety and compliance: checks and approvals before an agent acts.

  • Evaluation: testing agent behaviour before it ships and while it runs.

  • Working practices for teams that run agentic systems day to day.

Insights
LongMemEval results: the memory record beats full context on every model, +28 points at 19× lower cost on deepseek-v4-pro

Research note · August 2026

Memory over context: accurate, low-cost long-term recall for LLM agents.

A compact record of dated facts, read by a cheap model with three small tools, beats the full transcript on every model tested, at six to nineteen times lower cost. And it never invents an answer.

Read the research note

Start with a conversation.

Thirty minutes on an insight call with Jake or Ben. You leave with a clear view on whether a Working Week is worthwhile, and a fixed price in writing if it is.

Book an insight call

Next opening · Wed 7 Oct, 9:00 am AEDT

Jake from Hivemind discussing an AI solution with clients