Month-end is full of jobs that eat time without needing much thought. In this first fireside, Alex explains how Translucent's agents work: they sit on top of your accounting data, read and reason over it, and only act once you've approved the next step.
Oli then runs two jobs live on a client file: a prepayments run built from the company's own Excel schedule, and a supplier statement reconciliation from a PDF, finishing with journals posted back to Xero.
Key takeaways
- Translucent connects directly to Xero or QuickBooks and reads transaction-level detail, not just report totals: creation and posting dates, coding and tax-rate changes, descriptions and attachments.
- Upload the Excel templates and schedules you already use. The agent reads them and returns its work in the same format.
- Prepayments: it ties the opening balance to your uploaded schedule, finds new prepayment additions in the P&L and the prepayment account, calculates the release schedule and ties it back to Xero.
- Supplier statements: upload the statement as a PDF and the agent reconciles each invoice against the ledger, then gives you an Excel reconciliation to review.
- Journals go through an approval step before they're posted, with a two-way sync so Xero and Translucent stay up to date.
- Every job is a template you can tailor per client or for your whole practice, and jobs can be grouped into the checklists you already run each month.
- Your team can see every job in progress and pick up or take over someone else's work.
Read the transcript
Alex: Hello everyone, and thank you for joining. We're recording this and will share it afterwards, so there's no need to take notes. We've kept it short and focused, about half an hour. There's a Q&A box on the right, and Deane, Oli and I will answer questions as we go.
Alex: The aim today is to show you Translucent. Very simply, Translucent is the AI bookkeeping assistant for accounting and bookkeeping practices and in-house finance teams. You choose which jobs to hand over and give it supporting documents in whatever format you work in each month. It reads and understands those documents, follows your instructions, comes back for clarification when it needs it, and asks for sign-off before it does anything.
Alex: There's plenty of talk about AI in accounting, but very few people are showing what their tools actually do. So today we want to lift the bonnet. We'll cover where Translucent came from, give you an overview of how the platform works, and then Oli will run a live demo.
Alex: We started in February 2022, solving the multi-entity and group accounting problem: working across many Xero or QuickBooks files and trying to get one view of what's happening. To do that, we built a platform that brings in all the important information from your accounting software, and our first products used it for consolidation, intercompany and treasury. We've served over 15,000 businesses along the way, and those products still run today.
Alex: Practices and businesses kept asking if we could do more with their data. When AI and large language models changed what was possible, we could finally use that data the way we'd always intended. So we spent the last twelve months rebuilding the platform from the ground up as an agentic tool, so accountants and bookkeepers can hand over more of the work they repeat each month. We announced it in June.
Alex: The goal is that Translucent does the manual work, so you have more time for judgement and the work you trained for. A lot of each month goes on collecting data, reviewing coding and tax treatment, preparing schedules and rolling balances forward before the real work can start. With Translucent you start a job, and soon you'll be able to schedule jobs, and the agents collect the relevant transactions, do the review and analysis, reconcile your documents against the ledger, draft journals and invoices, and present the work for you to review. You come to jobs that are already well under way.
Alex: How do the agents work? If you've used Claude, ChatGPT, Gemini or Copilot, that's the foundation: they read and understand information, reason about the task they've been given, and ask for input before they act. The difference is that Translucent is built for finance. The jobs you do every month are built into the system, so the agents have a framework to work within.
Alex: The agents sit directly on top of your accounting data, whether you're a practice with hundreds of clients or a business with one file or a whole group. They read individual transactions, not just the trial balance: the creation date of a journal against its posting date, how suppliers are coded, changes to account codes or tax rates, descriptions, and the attachments themselves. They reason over that the way you would, for example recognising that an annual contract probably needs to be spread over twelve months, or that a supplier's goods need a particular tax treatment. And they only act with your guidance.
Alex: Your policies matter most. Translucent can't work in a vacuum. You give it your approach, the nuances of each client file, the Excel templates you use for working papers, schedules and supplier reconciliations, and your rules on materiality and frequency. That's the framework the agent works inside.
Alex: Most importantly, all of this works with your team. Your team has full control over what the agent does, full visibility over the jobs that have started and need input, and control over which jobs run and how often.
Alex: Everything is organised into jobs: a defined task done within a period, such as reviewing manual journals, preparing a template for a client call, or your monthly prepayments. We've built templates for these using standard accounting processes, and you can edit them per client or for your practice. You can then group jobs into a checklist, the same checklists you already run every month. Now let's see it working.
Oli: This is where you land. It shows all the jobs you have running. You set the agent off, and it tells you when it needs a human to review something. You can pick any accounting task from here.
Oli: We'll run prepayments for one client for last month. One of our founding principles was to work the way you already work, so I'll upload the company's standard prepayment schedule. The agent reads it, understands what's in it, does the prepayments and fills it in. I'll set a deadline so there's a record of when it's due, and start the job. The agent sets out a plan and works through it. Reading the workbook and planning takes a little time, so while that runs, let me show you a completed job.
Oli: This is a supplier statement reconciliation. You've received a supplier statement as a PDF and uploaded it here, and you can view it in the platform. The agent can read anything you upload, and it always has access to your accounting data, so with those two it reconciles the statement to your ledger and produces a supplier statement reconciliation for you to review.
Oli: I wanted it to work in Excel because you can upload all your templates for each client or company, and the agent always uses them. It's like onboarding someone new: here are the templates and the work I want done. In the output, each invoice is tied back to both the ledger and the statement. When you're happy with the work, mark the job as done and it leaves your list.
Oli: Back to prepayments: it's finished. It tied the opening balance to the uploaded schedule, looked through the P&L and the prepayment account for new additions, calculated the release schedule and added it to the template, tied back to Xero. You can download the workbook for your client file. To post the journals, there's an approval step, and with the two-way sync both Xero and the platform are updated.
Oli: You can run jobs across every company or client. This view is filtered to my jobs, but you can check any other company, see your colleagues' work and take it over if you need to help or move something forward.
Alex: The key point is that it's collaborative. You drop in your Excel files, it works from them and rolls balances forward. Today we focused on a few jobs: your prepayment schedule, reconciling balances, and draft journals you can post to Xero.
Alex: We'd put everything in two buckets. First, analysis and reconciliation, which language models are naturally good at: reading, comparing and spotting what doesn't tie. We've built the structure so that when you ask it to analyse a specific account, it pulls exactly that account, and when it reconciles a supplier statement it knows which details matter most. Second, doing the work: the jobs that repeat every month and take time but not much thought, like reconciliations or finding prepayment and accrual additions. That's where the biggest opportunity is. You can start several agents across several clients and come back to work that's already under way.
Alex: We're now building larger workflows with design partners: reviewing a new client take-on to scope the work, month-end checklists, year-end review and tax prep. And we'll run these firesides every three weeks to show what's new and hear what you want next. If there's something you'd like us to show next time, let us know. Thank you all for joining.