What’s new in the Toolbox.

A new table on My Expenses shows every submitted transaction that someone below you is assigned to approve — your direct reports’ reports, all the way down your chain. It’s formatted exactly like your other dashboard tables: sort by any column, search, click a row to open it, or hit expand for the full filterable view with an employee filter and a running total.

Open any team transaction and you can do everything the assigned approver can — Approve, Reject, Edit the classification, classify on someone’s behalf, even Combine or uncombine. There’s no reduced, read-only mode for managers: if your report can act on it, so can you.

This is oversight, not another inbox. Your team’s transactions never add to your To Approve count, never trigger a notification, and never show up in your daily digest. Those still belong to the assigned approver. The new table is simply there when you want to look or step in.

Your own direct reports work exactly as before — their submissions stay in your To Approve queue and keep driving your alerts. The team table only covers the levels below that, so the two never overlap.

Every recommendation now has Approve and Reject buttons, each with a box for your rationale. It’s entirely optional to act — if you do nothing, the recommendation just stays as Remi’s standing suggestion. There’s no “defer” to manage.

Decisions are available to Operations Leads and Admins, and the decision itself is instant — no AI call, just a save.

An approved or rejected recommendation is now preserved across syncs, along with its rationale and who decided it — so the call stays visible even when there’s no fresh recommendation. It’s only replaced when something truly changes: the recommended rent range moving beyond a set threshold (default $25 or 2%). Small data wiggles between syncs no longer wipe what you decided.

Every decision is also captured in Remi’s Action Log with the rationale, the person, and the date.

After you decide, Remi distills your rationales into durable notes for that unit type — the recurring reasons you accept or reject a direction, plus facts the data can’t see (“recently renovated,” “in-unit laundry,” “55+ community, low turnover,” “these comps are stale”). Those notes are fed into future recommendations and updated as you keep deciding — anything a newer decision contradicts gets dropped.

Remi shapes its reasoning from what it learns — it never changes the pricing math. The numbers stay deterministic; your rationale informs the story around them.

Record an overnight stay right on the report — provider, dates, and amounts — and attach the itemized folio when you paid on a personal card. Get your business unit head’s preapproval before you travel, and you’re set.

Staying at a company-managed property? You don’t need to file a lodging expense — just record the approved overnight so your meal per diem is counted.

Reports approve section by section. Your mileage goes to your manager; lodging and meal per diem go to your business unit head. You’re notified as each section clears, and anything sent back comes with a specific reason so you can fix just that part and resubmit.

When you drag a receipt onto the classify page, the AI now reads individual line items, not just the total. If it finds items from different expense categories — maintenance supplies and a capital item on the same Home Depot run, for example — a violet card appears with a proposed split: amounts per line and a suggested GL account for each.

Click Apply suggested split and the form is pre-filled with one line per category. Amounts sum to the transaction total. You review the property and any missing fields, then submit. If the suggested split isn’t right, dismiss the card and classify normally.

The AI sees your organization’s actual GL accounts when it reads the receipt, so suggestions reference account numbers from your chart of accounts — not generic categories.

The amber suggestion panel previously showed “Used Nx before” for the top match. Now every suggestion shows a ratio — 14 of 15× — so you can see at a glance how consistent the pattern is. 14 of 15 means this vendor has landed in the same GL account at the same property 93% of the time. 3 of 12 is a weaker signal worth reviewing before applying.

Suggestions are also ranked by your own classification history first. If you’ve personally coded this vendor to a particular account multiple times, your pattern surfaces above the company-wide aggregate.

Any transaction whose total is $30 or less no longer requires a receipt or a “Reason for No Receipt” note. The field disappears entirely on small charges — you can still attach a receipt if you want one on file, but nothing’s required to save and submit.

The threshold is based on the original transaction total, not the per-line amount, so a split that sums to $25 also clears it. Approvers see the same rule — small purchases without a receipt shouldn’t be bounced back.

Gas station pre-auth plus the real charge. Hotel deposit plus the final folio. Restaurant bill plus the tip adjustment. Amazon orders that ship in two pieces. These all post as multiple card transactions but share a single receipt. Now there’s a Combine button right next to Split.

You classify and attach the receipt once; every combined transaction inherits the same property, GL account, vendor, description, and receipts. Submit, approve, return — they all act on the whole group together.

The Credit Card Settings page has three new sections. The AI Suggestion Confidence Levels card lets admins independently enable or disable High, Medium, and Low confidence suggestions. High is ≥80% consistent with ≥5 prior classifications; Medium is 40–79% or a short perfect history; Low is everything below that. Low confidence is off by default to reduce noise.

The Splits & Combines card adds on/off toggles for both features independently. Disable splits if your workflow doesn’t need them; disable combines if you prefer transactions to stay independent. Both default to on.

You fill up at Speedway on a Tuesday. The pump pre-authorizes $1.00, then settles to $43.17 the next day. Two card transactions arrive in your queue: a $1.00 pre-auth and a $43.17 final.

1. Classify the $43.17 transaction first (Speedway, Fuel & Auto GL, the property you were driving for, snap the photo of the pump receipt). Save & Next.
2. Re-open it. Click Combine with other transactions.
3. The $1.00 pre-auth appears in the picker. Tick it. Click Combine.
4. Done. Both transactions now share your Speedway receipt and your fuel classification. Submit sends them both up for approval; one Approve from your manager covers both.

A two-night hotel stay with a $200 deposit charged on April 28 and a $487.30 final folio on May 2. Both rows on the same statement. The $200 deposit by itself isn’t a meaningful business expense — it’s part of the lodging total.

1. Classify the $487.30 final folio with the full hotel context — property, GL account, attendees if it’s a meal-eligible trip, and attach the folio PDF.
2. Combine the $200 deposit into it. Both rows now share the folio.

Combining across statements isn’t supported. A combined group has to live inside a single statement period — if the deposit and final fall on either side of a statement close, classify the deposit on its own statement and the folio on the next.

Click Uncombine on any member transaction to dissolve the group. Every member keeps the classification it currently has — nothing is reset — but the rows go back to being independent.

Returning a combined transaction from the accounting page also dissolves the group automatically so you can re-classify pieces separately.

Combined transactions appear with the same green “Combined N/M” badge on the Approvals queue and the Accounting export. One Approve covers every member of the group; one Reject or Return sends every member back at once and dissolves the group, so each row can be re-classified on its own afterwards. Verify the shared receipt covers the full combined total before approving — that’s the only added review step.

The old per-tool sections are gone. Now every task sits in a single list with red, amber, and grey dots — the same urgency dots you see on the dashboard. A red dot means overdue, amber means due soon, grey is normal. Each row tells you the tool, the task, and how long it’s been waiting (“6 days,” “today”), and clicks straight into the next item to act on.

A status line at the top of the dashboard summarises tasks across your tools. Quiet when you’re caught up — and when you’re not, it ranks the work so the most pressing items are at the top.

• Urgency dots. Red means overdue — e.g. a credit-card transaction un-classified for more than 3 days, an approval sitting more than 2 days, or a debt note within 60 days of maturity. Amber means due soon. Grey means a normal item in your queue.
• How old it is. Each row carries a short age suffix (“5 days”) so you can see what’s been waiting on you longest.
• Click straight into it. Rows deep-link to the specific item to act on, not a list page — so “Classify next” really does take you to the next transaction.
• A red tint when you’re behind. The status line border goes red and shows a small “N overdue” tag the moment anything crosses its threshold — easy to see from across the screen.

Overdue thresholds are per tool: 3 days to classify a credit-card transaction, 2 days to approve one. Lowe’s mirrors those. Investor letters: 3 days in review. L10 to-dos: more than a week past the meeting they were added in. Remi rent recommendations: 14 days since they landed. Debt maturity alerts: within 60 days = overdue, within 90 = due soon. Required training is always overdue until completed.

Each row gets a small snooze button. Pick “Tomorrow,” “Next Monday,” or “In 1 week” and the row drops out of your status line and per-card count until then. Useful when an approval is genuinely blocked on something or you’ve already pinged the right person — quieter signal, less noise.

Snoozes are per user. Yours don’t change what your manager or teammate sees. Maximum 30 days per snooze; after that the row comes back so nothing falls off the radar forever.

The expanded status line now has a separate “Waiting on others” section for items you submitted that are sitting in someone else’s queue. You can’t clear them, but you can see them — handy when you want to nudge the right approver instead of guessing whether your submission landed.

The header count and per-card amber pip only include items you can actually act on. “X tasks waiting” means what it says — the “others” bucket is below the fold.

A small green “N cleared this week” tag appears next to the summary count, totaling items you actioned in the last 7 days across every tool. Approvals you signed off, transactions you classified, to-dos you closed, courses you finished — they all contribute.

Counts your work across CC, Lowe’s, Travel, Investor Letters, Training, and Scorecards. Debt and Remi don’t contribute (no user-action timestamps to count) but the status line still stays honest about everything else.