My Work

FreshBooks wants AI that can eventually run parts of a small business — invoicing, follow-ups, financial insight — the work owners are already too busy to do themselves. But nobody hands a stranger the keys on day one.

The strategy reframes the problem away from what can the AI do and toward what has it earned the right to do. Two questions determine how fast trust can be built: is this action facing the client — high stakes, reputational risk, slow to earn — or happening inside the owner’s own system, where trust builds faster because the cost of a mistake is lower.

That produced Earned Autonomy, a four-rung ladder from Demonstrate to Delegate, mapped across two tracks — Relationship Intelligence (client-facing tone and communication) and Operational Intelligence (business structure and insight) — each rung tied to a specific emotional target: relief, then confidence, eventually freedom.

FreshBooks agents are positioned as the First Employee — something you onboard and trust incrementally, the way you would a new hire, not a black box you either trust completely or not at all.

The same communication architecture built for in-app messaging (see: In-App Communications) became the backbone for agent communication too — collapsing marketing messages, workflow nudges, and autonomous agent actions into one coherent voice instead of three competing ones.

And the strategy names its own risk plainly: you can’t ask someone to delegate work to an AI teammate while the system underneath it — the ledger, the reconciliation, the system of record — still feels unreliable. Trust has to be earned in the foundation first.

Every team had a message, and every message was a modal — trial reminders, billing nudges, feature launches, NPS surveys, all fighting for the same interruption. Customers said it plainly: it felt like being advertised to inside a tool they paid for.

The reframe: stop asking how do we get this message seen and start asking does this moment deserve an interruption. A fuzzy, subjective grid — engagement versus relevance — was replaced with an objective one: Criticality × Urgency.

That grid became a framework — AP/DC: Act, Prepare, Decide, Consider.

Act

High stakes, high urgency, allowed to interrupt. Involuntary churn, security alerts, “last day to submit payroll.”

Prepare

High stakes, lower urgency. Passive but visible, and built to escalate — a trial countdown that quietly becomes a blocking modal as the deadline closes in.

Decide

Optional, time-boxed. Upgrade prompts, beta invites, NPS asks.

Consider

Everything else. Release notes, achievements, partner news. Present, never in the way.

The framework produced its own UI vocabulary: Hotspot, a small badge marking new information without breaking flow, and Hotspot + Contextual Modal, a banner that escalates once, then disappears for good once it’s been seen.

The payoff nobody planned for: two years later, when FreshBooks needed a communication system for its AI agents, the same AP/DC framework absorbed it without a redesign — a two-way “Agentic Collaboration” channel sitting right alongside the original one-way “Workflow Communication” channel. One system, still standing, still legible.

FreshBooks’ native mobile apps dated to 2016 and had been in keep-the-lights-on mode since 2023. Real App Store reviews said it plainly — one star: “Buggy mess… invoices don’t load properly… gets stuck in a loop.” Zero stars, from someone who said the app cost them money outright: “Loses my mileage every single week… lost so much money in write offs.”

It mattered too much to leave broken. 45% of new signups first log into FreshBooks on mobile. Roughly a quarter of all active users were there monthly, generating 8.5% of paid accounts and 12% of new trials — and eight of the company’s top fifty support tickets traced straight back to it.

The first attempt was FreshBooks Connect, a React Native prototype aimed narrowly at payroll and field-service crews. Discovery showed the bet was too narrow to justify migrating the whole app off native — so it was pulled from the app stores rather than shipped halfway. That’s the harder, less comfortable case-study beat, and it’s the one worth keeping: killing a narrow bet instead of shipping it anyway.

The pivot organized the rebuild around two modes an owner actually moves between, rather than mirroring the web app’s feature list — Modes, not roles.

Serving Mode

Delivering the work: invoices, estimates, payments, time, mileage. Fast, one-handed, in the field.

Manager Mode

Running the business: financials, team, the accountant relationship.

Design started at the overlap between them — Quick Actions — instead of trying to serve every role and vertical at once.

The grounding was usage data across roughly 45,000 monthly active users (invoicing and expenses alone were three-quarters of all mobile engagement) plus interviews across nine small businesses — a landscaper, a cybersecurity firm, a legal practice — to stress-test the mode split against real working days.

Mapped against Xero, QuickBooks, and Zoho on two axes — quick actions versus full platform, built for accountants versus built for owners — the gap was clear: fast, owner-first, not accountant-mediated. Nobody else was sitting there.