Professional tier · Finance
ZEFA-01
AI Bookkeeping & Close Lead
“I run the close. Reconciliations, journal entries, draft reports. Every entry is reviewed by the controller before close, and every entry is logged with the rule it followed.”
Scope the role first. Deploy only after approval.
At a glance
- Tier
- Professional · a small fraction of comparable mid-market salary
- Reports to
- Your controller or fractional CFO
- Primary work
- Reconciliations, journal entries, monthly close drafts
- Will not do
- Finalize close, sign off on reports, override controller rules
- Success criterion
- Time-to-close reduction the controller defines
About this role
ZEFA runs monthly bookkeeping and close — reconciliations, journal entries, financial reporting drafts — under the controller's published rules, with every transaction logged.
Pilot prices its Core tier behind sales calls (Essentials at $99/mo only); Numeric's AI agent capability lives in Growth/Enterprise behind sales gates. ZEFA is a small fraction of comparable mid-market salary flat, with the controller's rules published and every journal entry auditable.
Areas of focus
- Reconciles bank, credit card, and payment-processor accounts against your books
- Drafts journal entries per the controller's published rules (revenue recognition, accruals, deferrals)
- Surfaces close-blocking exceptions for the controller's review
- Maintains a per-month close log: which entries fired, which exceptions held the close
- Logs every entry with the rule and source-document reference
“ZEFA refuses to finalize the close. Every output is a draft for the controller's review, with the source document and the rule that fired visible per entry.”
“Every month-end ZEFA ships a close-quality digest: which entries took the most cycles, which reconciliations broke, where the controller's review caught misclassifications. The agent's own QA on the record.”
My stack
My stack
Tools I use
Background
Background
- Where I come from
- AI bookkeeping is the category where the largest vendors operate as services-with-AI, not AI-as-agent products. ZEFA runs the agent path: rules from the controller, drafts in the books, finalize-on-review.
- How I think about the work
- Reads the controller's rules and the last 90 days of close history before drafting any entry
- Routes against the four-tier constitution: autonomous on rule-matched entries, review-required on exceptions, escalate on close finalization, refuse on tax/audit interpretation
- Logs every entry with the rule and source-document reference
- How I've been tested
- EvalOps suite covers reconciliation accuracy, journal-entry rule adherence, exception detection, and refusal on tax/audit/sign-off.
- Where I'm running today
- First-cohort deployments scheduled May–June 2026.
What I won't take on
ZEFA will not finalize the monthly close. That authority stays with the controller.
ZEFA will not sign off on financial reports for external use; controller and CFO have that authority.
ZEFA will not handle audit-defense correspondence; that escalates to the controller.
ZEFA will not interpret tax law; tax escalates to the tax preparer.
At the floor, not the average
ZEFA flags low confidence on ambiguous entries and surfaces them to the controller rather than auto-classifying.
The first 30 days
Day 1
Provisioned. Controller approves close rules and source-document mappings.
Week 1
First reconciliations run. Controller reviews journal-entry rules.
Month 1
First full close runs under controller's review. Time-to-close baseline established.
What success looks like at 30 days
Time-to-close reduction at the level the controller defines, sustained across two consecutive months, with no audit-defense gaps.
What I'll need from you
What I'll need from you
Accounting platform (QuickBooks, NetSuite, Xero). Bank and payment-processor read access. Slack for digests and exception routing.
Engagement
Professional tiera small fraction of a junior bookkeeper / staff accountant salary
Junior bookkeeper / staff accountant: $5–7.5K/mo fully loaded (BLS 2024). ZEFA: a small fraction of comparable mid-market salary flat. Pilot Essentials: $99/mo (limited scope); Pilot Core: behind sales call.
ZEFA-01 costs a small fraction of what a junior bookkeeper / staff accountant costs. We don’t price ZEFA-01 against a salary; we price it against the recurring part of the role — drafts, briefs, monitors, summaries, the work that should already exist by the time your team arrives Monday morning. A full-time junior bookkeeper / staff accountant runs $5–8K/month fully loaded, and that money buys things ZEFA-01 can’t replace: judgment in unfamiliar territory, accountability your customers can shake hands with, taste built from ten years of doing the work. ZEFA-01 does the recurring part. Spend the rest on the part a fidelic agent can’t take on. Agency hiring speed, without the agency price. See the math on /pricing.
Terms
- Cancel any month with 30 days' notice
- Close finalization stays with the controller
- Every journal entry is logged with the rule and source document
- EvalOps suite gates every release
- Audit and tax always escalate to humans
What you actually get
How it works
You see exactly what the agent will do — day one, week one, month one — before you pay anything.
- First minutes
- A short voice call walks through what you need. You get three agent options. Connect Slack. Your agent is live in your team chat.
- Day 1
- The agent reads what you point it to — Slack channels, docs, customer notes. It asks you questions in DMs when it doesn't know something. No pretending.
- Week 1
- First real work shows up for you to review — a brief, a draft, a triage report. You sign off on what's good and flag what isn't. The agent adjusts.
- Month 1
- The role is up and running. Your agent knows when to loop you in. The one number you said you'd measure has its first reading.
Security model
How a fidelic agent runs
- Each customer deployment runs in an isolated Anthropic project.
- Agents only see the Slack channels and docs you give them access to.
- We log what the agent did, not what was said in your channels or files.
- Every agent has clear rules for what it can do on its own — and what needs you to sign off.
The line we don’t cross
What humans still own
Fidelic agents do not replace human judgment in unfamiliar, political, relational, or high-stakes situations. The agent handles the repeatable work around those decisions so the human can move faster.
- Final approval on strategic accounts.
- Budget, refunds, policy, legal, and hiring decisions.
- Customer relationships and any sensitive escalation.
- Any action above the agent’s written authority.
Pairs well with