If you run finance for an import-export business, you already know the real bottleneck isn't invoices. It's the cards. Freight forwarders, customs brokers, supplier deposits, the duty payments that hit your corporate card at 2am local time in three currencies. So when teams ask me how to build AI agents for card workflows that actually keep up with cross-border spend, my answer is usually the same: start by getting off the tool that's making you babysit every transaction. For a lot of importers, that tool is Docyt. This guide walks through moving from Docyt to the Aiinak AI Finance Agent in a typical one-to-two-week window, what breaks, and what you'll genuinely miss.
I've managed operations for 15+ years, and I've now run AI finance agents in three businesses. This isn't theory.
Why Import-Export Teams Build AI Agents for Card Workflows First
Here's the thing about import-export: your card statements are a mess by design. A single shipment from Shenzhen might generate a freight charge in USD, a customs clearance fee in EUR, a port handling charge in your home currency, and a fuel surcharge that posts two weeks later. Docyt does a decent job pulling those in and matching receipts. But it stops at categorization. Someone on your team still decides which shipment each charge belongs to, whether the FX rate matches your landed-cost model, and whether that $340 "handling fee" is a duplicate.
An AI finance agent is different because it acts, not just files. When you build AI agents for card workflows, the agent reads the card feed, matches each transaction to an open purchase order or shipment reference, flags the duplicate before it's paid, and books it to the right cost center — autonomously. In my experience deploying agents, that single shift cuts the manual touch on card transactions by something like 60-70% for a typical importer. Not because the AI is magic, but because it stops the human ping-pong of "which job is this for?"
The mistake most teams make is treating the migration as a software swap. It's not. You're moving from a tool that organizes work to an agent that does the work. Plan accordingly.
Week One, Days 1-3: Plan and Audit Before You Touch Docyt
Do not export anything yet. Spend the first three days auditing what Docyt is actually doing for you, because you'll be surprised how much undocumented logic lives in one person's head.
Make a list of every card workflow you depend on. For most import-export operations that includes: corporate card reconciliation, vendor invoice matching, multi-currency expense categorization, customs and duty tracking, and the monthly close. Write down which of these Docyt automates versus which your bookkeeper does manually inside Docyt. (This second list is usually longer than people expect.)
Then pull three things:
- Your chart of accounts — exactly as it sits in QuickBooks, Xero, or Sage. Aiinak's AI Finance Agent integrates with all three, so your COA is the spine of the whole migration.
- Your category and vendor mapping rules — every "if this vendor, then this account" rule you've built in Docyt over the years.
- A list of recurring exceptions — the weird transactions someone always has to fix by hand. Duty deferrals, partial supplier refunds, FX adjustments.
That exceptions list is gold. It's the actual test suite for whether your new agent is working. Most migration failures I've seen come from skipping this step and discovering in month two that the agent never knew about your bonded-warehouse handling.
Week One, Days 4-7: Data Migration Without Losing History
Good news first: you probably don't need to migrate transactional history out of Docyt at all. Your accounting system — QuickBooks, Xero, or Sage — is your system of record. Docyt was the processing layer on top. So the migration is mostly about connecting the Aiinak AI Finance Agent to your accounting platform and your card feeds, then teaching it your rules.
Connect the agent to your accounting system first and let it ingest the existing chart of accounts and vendor list. Then connect your corporate card feed and bank accounts for reconciliation. This is where Aiinak's bank reconciliation and accounts payable automation start doing real work — the agent begins matching live transactions against open POs immediately.
Now feed it your rules. Don't try to replicate every Docyt rule one-to-one. Instead, hand the agent your vendor mapping and your exceptions list and let it propose categorizations on a sample of last month's card transactions. Review what it gets right and where it stumbles. In my experience, an agent will nail 80-85% of card categorizations out of the gate using your existing data, and the remaining 15-20% is where you invest your training time.
One honest caveat: if you have years of receipts and documents stored inside Docyt, export those before you cancel. Docyt holds the source images. Pull a full document export and drop it into Aiinak Drive (the RAG search makes old receipts findable later), or your own archive. Losing receipt images is the one piece of history you can't reconstruct, and customs audits in this industry can look back years.
Week Two: Train the Team and Run in Parallel
Run both systems side by side for the first week of live operation. I can't stress this enough. Don't cut over cold.
Parallel running means your card transactions flow through the Aiinak agent and your existing Docyt process at the same time, and someone compares the output daily. It feels redundant. It is redundant. That's the point — you're verifying the agent books the same FX charge to the same shipment your bookkeeper would, before you trust it alone. A week of overlap is cheap insurance against a month of cleanup.
For training, the shift for your team is psychological more than technical. Your bookkeeper goes from doing categorization to reviewing the agent's decisions. That's a real change in the job, and people get nervous. Frame it honestly: the agent handles the volume, your person handles the judgment calls and the exceptions. The duty deferrals, the disputed supplier charges, the FX adjustments that need a human eye — those stay human.
Practical training steps that work:
- Daily exception review. Have your finance person spend 20 minutes each morning reviewing what the agent flagged. This builds trust fast and trains the agent on edge cases.
- Set budget alerts early. Aiinak's budget monitoring will ping you when card spend on a category spikes. Configure these in week two so they're tuned by go-live.
- Document the handoff. Write down which decisions the agent owns and which a human approves. Ambiguity here causes double-work.
Based on industry benchmarks for finance automation rollouts, teams typically reach comfortable trust with an AI agent within two to four weeks of live use. Don't expect blind confidence on day one. You shouldn't have it.
Go-Live, Common Pitfalls, and What You'll Miss From Docyt
Once your parallel week shows the agent matching your manual output consistently, cut over and turn off the Docyt processing. Keep your accounting system exactly as it was — nothing changes there.
Now the honest part. Here's what you'll actually miss from Docyt, and how the Aiinak AI Finance Agent compensates:
- Docyt's receipt-capture mobile app. If your buyers snap photos of paper receipts on the road, that workflow is genuinely good in Docyt. Aiinak handles document ingestion through email forwarding and Drive upload, which works but is a slightly different habit. Set expectations with your travelers.
- Docyt's bookkeeping-service feel. Some teams use Docyt almost like a managed service. The Aiinak agent is autonomous, not a human bookkeeping team — if you wanted someone to call, that's a real difference. The tradeoff is the agent runs 24/7 and doesn't have a backlog.
- Familiar reporting layouts. Your monthly reports will look different. Aiinak generates financial reports and real-time insights on its own schedule, which is more current but takes a cycle to get used to.
What you gain: the agent does accounts payable and receivable, reconciliation, and reporting without waiting for a human to start the task. For import-export specifically, the real-time multi-currency view is the thing people don't go back from.
Common pitfalls to avoid: don't migrate during your busiest shipping season (the parallel week needs attention), don't skip the exceptions list, and don't cancel Docyt before exporting your receipt images. I've watched all three of these bite teams.
The Cost Math: AI Finance Agent vs Bookkeeper
Let's talk numbers, because this is usually the real reason teams move. Aiinak's AI Finance Agent starts at $499/month. A part-time bookkeeper handling import-export complexity runs anywhere from $2,500 to $5,000+ a month depending on your region and transaction volume, and a full-time finance hire is well north of that. Docyt's processing tiers typically land in the few-hundred-to-low-thousands range depending on document volume.
So on a pure ai vs bookkeeper cost comparison, the agent is a fraction of a salary. But don't buy on price alone. The honest framing: if your card and invoice volume is low and stable, a good bookkeeper with Docyt is fine. The agent earns its keep when volume is high, multi-currency, and exception-heavy — which describes most growing importers. That's exactly the profile where automated financial reporting and autonomous reconciliation save the most hours, typically in the range of 30-50% of a finance person's week reclaimed for higher-value work.
If you're ready to build AI agents for card workflows that match your shipments instead of just filing your receipts, you can Deploy Finance Agent in an afternoon and start a parallel-run test this week: https://admin.aiinak.com/ai-agents. Start with one card feed, run it alongside Docyt for a week, and let the matching results make the decision for you. That's the lowest-risk way I know to find out if an AI finance agent fits your operation — without betting your monthly close on it.
Originally published on Aiinak Blog. Aiinak is an AI agent platform that runs your entire business — deploy autonomous agents for Sales, HR, Support, Finance, and IT Ops.
