RESEARCH & RESOURCES

Research, papers, and what we’re building

Bizer is built on a simple belief: small businesses deserve advisory-grade help, made practical by AI. Here’s our published work, the questions we’re actively investigating, and what’s shipping next.

PAPERS & OPEN SOURCE

Published work

Grounded by Construction: AI Agent Never Fabricates Your Numbers

How Bizer’s architecture keeps an AI agent grounded in your real data, so it never invents the figures your business runs on.

Whitepaper PDF · Video Podcast

A Cognitive Architecture for AI Agent Memory

The memory architecture behind Miles, how an AI sidekick remembers context across your business, week over week.

Whitepaper PDF · Video Podcast

BizerBrain on GitHub: our open-source AI agent memory project.

RESEARCH AGENDA

What we’re researching

The questions we’re actively investigating, grouped by theme, and open by design.

AI adoption friction · In progress

What stops small businesses from adopting AI, and what removes the friction?

Map the real barriers, trust, time, cost, technical confidence, and test which interventions most increase adoption among non-technical owners.

The approval gate and trust · Exploring

Does an “approval gate” increase owner trust and AI usage?

Measure whether requiring explicit human approval for every outbound AI action changes how much owners trust, and actually use, an AI sidekick.

Predicting cash crunches · In progress

Which leading indicators predict small business cash crunches 30 days out?

Identify the receivables, burn, and seasonality signals that most reliably forecast a cash shortfall before it happens.

AI invoicing and days sales outstanding · Planned

What is the impact of AI-assisted invoicing on days-sales-outstanding?

Test whether AI-drafted, approval-gated payment reminders measurably reduce how long invoices stay unpaid.

ROI of AI market research · Planned

What is the ROI of AI-curated market research for solo founders?

Evaluate whether founders who use an AI sidekick for market and competitor research reach validated positioning faster.

AI-assisted review responses · Exploring

Does responding to reviews with AI assistance lift local conversion?

Test whether faster, AI-assisted (human-approved) responses to Google Business Profile reviews improve local search visibility and conversion.

GTD plus an AI sidekick · Planned

How much time is recovered when GTD and an AI sidekick run the weekly review?

Quantify the time and follow-through gains when owners pair a weekly review with an AI that remembers context.

Continuous support and business outcomes · Exploring

What happens to outcomes when the support between advising sessions is continuous?
Investigate whether the assess–focus–act loop, running daily in software with humans in the loop, improves the results owners get from the counseling and mentoring they already receive.

PRODUCTS

Released

Miles Realtime Voice Chat: talk to Miles live, by voice.
Miles Biz Researcher, alpha release, miles.bizer.ai
BizerBrain, alpha release, github.com/kelsi-bizer/bizerbrain

ROADMAP

In progress

Miles, an AI agent for small businesses
BizerOS, an operating system for small businesses
Bizer Chat, a team chat for small businesses
BizerGPT, an AI chatbot platform for small businesses
BizerOS Appstore: the curated app store for BizerOS

FROM THE BLOG

Latest posts

1 week ago

Dual-Signal Retrieval for AI Agent Memory

A white paper on ranking an AI agent’s memories by use without making recall worse. A two-year simulation through production code shows why use and correctness must be two separate signals, why every ranking nudge must stay smaller than the gaps it reorders, and why recall and long-term memory need different clocks.

2 weeks ago

The Hippocampal Index for AI Agents

A white paper on why AI agents get slower and costlier the more they remember, and a fix borrowed from the brain: a small index of what the agent knows, kept current by every write and repaired offline. Research, reference design, use cases and an evaluation method.

2 weeks ago

Miles Now Keeps Track of Your Long Conversations

Miles, Bizer’s AI agent, now keeps a short list of what scrolled out of a long conversation, each line pointing to where he saved the details.

2 weeks ago

What Human Memory Teaches AI Agents About Long Conversations

AI agents lose the middle of long conversations. Human memory points to a better fix than a bigger window: an index of what scrolled away, each line pointing to where the details are stored.

Have a question we should be studying?

We’re always looking for sharp questions from owners, advisors, and researchers.

Get in touch