What Got Shipped - 1 & 2/9/26 - Cash In lands on the cash flow, an approval can be handed over, and your Word file becomes the template

2 September 2026 Two days in one entry. Cash flow stopped being a cost-only picture, approvals stopped stalling when someone goes on leave, action items started chasing themselves, and the document template engine now takes your own Word file as the source of truth. Features Cash In lands on the cash flow: money out, money in, and the net line between them "I can see what this project costs me month by month. What I cannot see is when the money actually arrives, so I have no idea how deep into my own pocket this job goes before it comes back." (Commercial Manager) Before. The cash flow was a cost-only view. Cost Out had a curve, a chart and a pinned band of totals. Cash In did not exist. Working out when a project was underwater meant exporting the forecast, pulling the claim schedule out of the head contract, and building the net position by hand in a spreadsheet, then doing it again next month because nothing carried forward. After. The cash flow now carries the full picture in one grid. Cash In rows are seeded from the Cost Out curve so you never start from a blank sheet, then split per contract into base and variation streams with a collapsed rollup, so a job with three contracts and a run of approved variations reads cleanly instead of as forty rows. Every future-month Cash In cell is editable. Click, type, save. The months you did not touch re-flow so the row total still equals what is claimable on that stream, the edited cell goes bold, and a guard rejects any override that would push a stream past its claimable value, telling you the overage rather than silently accepting it. Past months are certified fact and offer no edit at all. Overrides are stored per project, per forecast period, per contract, per stream and per month, so they survive a reload, stay inside the period they were made in, and carry into the next forecast period rather than evaporating. The Net Position section sits below, with a cumulative row and an opening position that folds in everything before the visible window, so the curve starts where the project actually starts rather than at zero. The chart is now three series. Cost Out in faded red, Cash In in faded green, Net Position as a thick blue line on its own right-hand axis with a zero baseline, and the fill turns red where the cumulative net goes under. The Cumulative and Monthly toggle moves all three together, and the pinned band above the grid mirrors exactly what the chart plots, four rows, so it no longer eats half a laptop screen. Section detail totals moved to their own section footers where they scroll with the rows they summarise. Below the chart, an insight callout says in words what the curve shows: how much the project has to fund out of its own pocket, the month that peaks, and the month it comes back. It detects every underwater run, not just the first, because retention alone guarantees a second dip, withheld monthly and released in a lump at completion. The deepest run is the headline, the rest are counted and expandable. It reads the cumulative curve whichever chart mode you are in, because it states a fact about the project rather than a property of the current view. And an approved cash flow is now reproducible. Approving a forecast freezes the inputs the grid depends on, the programme task windows, the sparse monthly actuals and the override cells, and recomputes the series at read time. Reopen last month’s approved forecast after a schedule change or a newly certified claim and you see what was signed off, not a quietly rewritten version of it. Impact. The two dates a builder actually needs, peak funding requirement and break-even, now come off the platform instead of a spreadsheet, and they hold their value once approved. On a mid-flight job that is the difference between finding out you are $1.2M underwater in February from your own forecast, and finding out from your bank. Known limitations. Payment terms and forecast-scoped retention join the same grid in the next slices, so today’s Cash In assumes cash arrives in the month it is certified rather than shifted by the contract’s payment terms. The Cash Flow tab on a live project. The pinned band at the top carries the same four series the chart plots, Cost Out, Cash In, Net and Cumulative Net, and the callout above it states the funding requirement in plain words: $131,933 at the end of the window and still falling, rather than leaving you to read it off a curve. Delegate a single approval, or hand your whole queue over while you are away "I am on leave for a fortnight. Every approval that routes to me is going to sit there, and the two that matter are the ones holding up a subcontract execution." (Project Manager) Before. An approval routed to you and stayed there. Going on leave meant approvals piling up, workflows stalling, and someone chasing you on the phone from site. There was no way to say "I am not the right person for this one" without an admin editing the workflow, and no record of who was originally meant to act. After. Two ways to move an approval, and both leave an audit trail. Out of office lives in the Delegations tab. Set a date range and a named delegate, and during that window every approval that would route to you routes to them instead. Notifications follow the delegate, and you stop getting reminder emails for things you cannot act on. You can create, edit or cancel your own out of office, and so can an org admin. Ad hoc delegation is a fourth action on any active step assigned to you, alongside Approve, Reject and Request Info. Pick a colleague, give a reason (mandatory), and that step reassigns. Later steps in the same workflow still route per the template, so one handoff does not rewrite the governance. Every decision now records the original assignee, who actually responded, the route it took (direct, out of office, or ad hoc) and the reason if it was delegated. That is queryable per record for a full audit trail, and per user, so you can see what you delegated and what was delegated to you. The system validates that a delegate can actually act on the record type before the handoff goes through, and it warns on delegation cycles. Impact. Approvals stop being a single point of failure attached to one person’s calendar. For a builder running Delegation of Authority properly, this is the piece that makes it survive annual leave, sick days and site absence without either stalling the job or quietly breaching the authority matrix. Action Items chase themselves "The action was assigned three weeks ago, it is due Monday, and nobody has touched it. I only found out because I opened the register on Friday afternoon." (Submission Owner) Before. An action item had a due date and nothing else. Nothing reminded the assignee, nothing escalated when it went past due, and the only way to know an action was drifting was for someone to open the register and read it. After. A daily sweep now runs in production across three reminder tiers: a lead reminder ahead of the due date, a reminder on the due date, and escalating overdue reminders.The lead tier is computed in working days, three of them, which matters more than it sounds. Three calendar days back from a Monday deadline lands on the previous Friday and leaves the assignee exactly one working day, which is the failure the whole feature exists to prevent, and construction deadlines land on Mondays constantly. So an action due Monday sends its lead reminder the preceding Wednesday, one due Thursday sends the preceding Monday, and a public holiday in between pushes it one working day earlier again. The sweep reads each project’s working calendar, country and state once per run, so a project that marks a weekday non-working has that reflected, and a project with no calendar configured still sends on a plain Monday-to-Friday basis rather than being skipped. Opening an action now shows a compact reminder-status block: the three tiers, each marked sent or not sent. That is the answer to the question people actually ask, which is not "did the system log something" but "has this person already been chased three times, in which case it is my turn to pick up the phone". The notification log always held that, but nobody opens the workspace notification log to answer a question about one action. Reminder emails also deep link straight to the action now, so the link in the email opens the action rather than dropping you on a register to go find it. Impact. Actions stop relying on someone remembering to look. On an HSEQ or submission workload where the cost of a missed action is a compliance gap rather than an awkward conversation, three working days of warning is the difference between a fix and an incident. KAI drafts your scope of works from the tender documents "I read the consultant’s spec and the drawings, then write the trade scope by hand, spelling out what is in, what is out and where the interfaces sit. It takes a day per package, and it is exactly where the gaps creep in." (Contract Administrator) Before. Per-trade scopes of work were written from scratch against the spec and drawing set for every bid package. Slow, and the classic failure mode is well known: the subbie priced structural steel but excluded shop drawings and cranage, nobody caught it, and it surfaced at install. Every subcontractor then priced a slightly different scope. After. A "Draft scope with KAI" action sits on the Scope step of bid package creation. Point KAI at the tender document set already in your project documents, spec and drawings, and it proposes a structured scope per trade: inclusions, exclusions and trade interfaces. The output drops into the existing Scope of Works structure as an editable draft, not a locked block. KAI suggests, you decide, and nothing publishes without your confirmation.Impact. The first pass at a trade scope goes from a day of reading to a review. More importantly, KAI reads the whole document set rather than the parts a busy CA got to, which is where the scope gaps live. Known limitations. No quantities, takeoff or measurement here, that is the 2D Takeoff line. Nothing auto-publishes. The Scope of Works library stops being copy-paste "I have a master scope in the library. Someone modified it on a project, and now I have no idea which projects are running the modified version, or whether the master is still what anyone actually uses." (Commercial Manager) Before. The library copied dumb text. A project-level scope that got modified lost all connection to the master it came from, so nothing could tell you which bids or subcontracts a master was attached to, or how far a project copy had drifted. Worse, a modified project scope that attached to a bid correctly failed to carry through to the drafted subcontract, so the scope you agreed at bid was not the scope in the contract. Impact. A scope library is only worth maintaining if you can see what depends on it. This is what turns it from a folder of text into an asset that carries from library to bid to subcontract intact. Your Word file becomes the template "Our subcontract has been formatted the way our clients expect it for fifteen years. I do not want to rebuild it in a form builder, I want to upload the document and tell you where the fields go." (Director) Before. Document templates were built and edited inside the platform. Every format change a customer wanted, a margin, a table layout, a clause block, came back to us. Customers with a document that already looked right had no way to bring it in as-is. After. A Word (.docx) file is now the template source, with variable injection and PDF generation handled on the backend.Upload a .docx and it lands straight on a full-page template workbench with a backend-rendered PDF preview, so what you see is what will generate. Two preview modes, Template and Simulated, let you check the raw variable layout or the document as it will read with real values. Tagging suggestions are drawn from your actual Company Settings values, and Find and Tag replaces literal text in the Word file with a variable without re-uploading. Live preview is now embedded in the Generate and Send Document flows, so you see the document before it goes out rather than generating and checking. Workspaces can create their own custom variables (a constant, a record field, or a prompt at generate time). And repeating table rows work: subcontract schedule-of-values lines and purchase order lines expand per row across all four document paths, the Generate modal, send document, subcontract packs and purchase order packs. A template section can also ask to stay on one page rather than splitting across a page break.Impact. The customer owns their own document. That is the unlock: a builder can fix their own formatting, add their own clause, and reload the template themselves, instead of raising a request and waiting on us. Known limitations. Format changes still go through download, edit in Word, replace. The in-browser editor and KAI conversational document editing are the next phase. A Home page that opens on your work "I log in and land on a project list. Then I go to four different registers to find out what is actually waiting on me today." (Project Manager) Before. Signing in dropped you on a project view. Whatever was assigned to you, pending your approval, due on the programme this week or addressed to you in correspondence lived in four separate registers, and finding it meant knowing where to look. After. A new Home view pulls your work into one place, with widgets for the actions assigned to you, the approvals pending on you, your programme tasks for the week, and the notices related to you. A view-as filter lets you switch the whole page between three perspectives rather than reading one fixed cut, and the layout now works properly on a phone or tablet. Impact. The first screen after login answers "what is waiting on me" instead of asking you to go and find out. Known limitations. This is a v1. The widget set and layout will grow. The new Home view. The four widgets are the actions assigned to you, the approvals pending on you, your programme tasks for the week and the notices related to you, with the view-as switch at the top right and your project portfolio down the left. Raise a clarification against a tender package, and count them on the register "Tender clarifications are part of the tender issue. They become the foundation for the procurement and subcontract packages. Right now they float loose with no link to the package they clarify." (Director) Before. A clarification notice knew its workspace and its project but not what created it, so nothing could answer "show me the clarifications on this tender package". The Tender Packages register carried a hard-coded zero for RFIs and clarifications because there was nothing real to count. After. A clarification notice can be raised directly from a tender package, and the link is set at creation and immutable once issued. The tender package detail view lists its clarifications with reference, status and due date. The clarification and RFI counts on the Tender Packages register are now real numbers. Clarifications issued against a package appear on the Tender Issue Sheet and travel with the set. Notices also gained an optional Category field, so a clarification is a typed thing rather than a free-text note.Impact. Clarifications stop being loose correspondence and become part of the tender record, which is what makes them usable as the foundation for the subcontract scope later. Improvements The Delegation of Authority goes back to one workspace setting, with a project override when you want one Before. The approvals rebuild made the authority ladder and workflows project-scoped by default, which meant every new project needed its own configuration before anything could route. It also meant the Highest Authority approver was invisible on every project scope, so it looked like it was not saving. After. The workspace-wide configuration is the primary surface again. Add a workflow or a threshold and it writes at workspace level, which is what almost every builder wants: one business rule per record type, resolved against either workspace roles or project-assigned roles. Project-level is now a deliberate Project Override you opt into per record type or per threshold ladder, with a confirmation that spells out that the workspace default stops applying to that project once you create one, and a revert that hands the project back to the workspace default. The Highest Authority approver now displays correctly on every scope, and a stale fetch that could clobber a project’s scope was fixed.Impact. Setting up approvals on a new project is nothing again, it inherits. And the builders who genuinely need different thresholds per project (different clients, different value bands) still get it, without imposing that complexity on everyone else. Provisional Sums read the same way everywhere in a claim Before. Provisional Sum presentation was only finished in the Create Claim modal. Edit the claim, open the claim detail, or send the PDF to the client, and the PS structure vanished. A claim built correctly lost its separation the moment anyone looked at it. After. Provisional Sums are now presented consistently across the whole claim lifecycle. PS lines are separated from the original contract schedule of values and from approved variations in Edit Claim and Claim Detail as well as Create Claim. Approved PS is claimable, pending PS is visible but disabled and reads zero. The claim PDF carries the same separation, and head contract claim output has a dedicated Provisional Sums summary row, because a provisional sum is not a variation, and neither AS 4000 nor NZS 3910 treats it as one. Impact. The claim your client reads matches the claim you built. The value was always right, the presentation was not, and on a client-facing PDF presentation is what causes the argument. Hide the summary cards on any register, and it stays hidden Before. Summary cards plus filters could leave a handful of table rows on a laptop. One customer filtered the Company Directory by trade and saw two companies. The Budget Register had a Hide summary toggle, but it was local to that page and forgot your choice on every reload. After. The toggle now exists on the Budget Register, Cost Forecasting, the head contract and subcontract claim, invoice, payment schedule and variation registers, bank guarantees, the Invoice Register, the PO Register and the Company Directory, and it remembers your choice per register, per user, across reloads. A hidden summary no longer flashes a row of skeleton cards on load either. Impact. On a laptop you get your rows back, and you only have to say so once. Contacts get their own view in the Company Directory Before. The Company Directory listed companies. Finding a person meant knowing which company they sat under and opening it. After. The Company Directory gains a Contacts view alongside Companies, a reverse lookup of every person at every company connected to your workspace, grouped by company. Contacts now carry workspace-owned extended detail behind an information icon, viewable and editable in place, and Trade and Region can be edited inline from the drawer with chips shown in the list. The Trade Specialties picker reads your workspace Trades registry rather than free text. Impact. The directory becomes a people directory as well as a company directory, which is how anyone actually searches it when they are trying to get a price. The Company Directory. The new Companies and Contacts switch sits at the top right, and the Hide summary link above the cards is the toggle that now exists on every register and remembers your choice per register. One Tender Packages page Before. Incoming tenders lived in two places, the Tender Register and the Tender Tracker, with overlapping capability and no clear answer to which one was authoritative. After. The remaining Tender Register capability has moved into the Tender Tracker so there is one incoming tender surface. The old page is retired and its URL still resolves, so saved links and bookmarks keep working. Impact. One place to look, and nobody has to remember which register a tender was entered into. The Tender Tracker loads fast and stops losing your place Before. The Tender Tracker took too long to load on a real workspace, and it would lose context and need a hard reload. After. The tender list fan-out was profiled and fixed at the data layer, and the context loss was fixed alongside it as the same root cause. Both were measured before and after on a full-sized workspace. Impact. The register you open twenty times a day opens. The progress claim uses one name per money column Before. The head contract progress claim showed the same five money columns twice, whole-of-contract in the Contract Summary and per line in the Progress Claim Breakdown, and used a different name for every one in each place. "When Complete" against "Contract Value". "Previously Approved" against "Previous". "Remaining" against "Balance to Finish". It also ordered them opposite ways round. A customer reading their own claim asked how "Remaining" differs from "Balance to Finish". It does not. After. The Contract Summary adopts the Breakdown’s names and its period-then-cumulative order. One vocabulary, one order, across the whole document. Impact. A client reading a claim stops asking what the difference is between two columns that were always the same number. A proposal can go out as a section summary Before. The proposal builder showed the full line-item detail, which is not always what a client should see at proposal stage. After. Quoted items can be imported by estimate section, and a proposal can present section-level sell rollups only, with the same section summary mode carried through to the client PDF. Impact. You can put a proposal in front of a client at the level of detail the conversation calls for, without stripping lines out by hand. Fixes An invoice priced GST-inclusive now converts at the right value Before. The invoice register detected whether a supplier’s line amounts were GST-inclusive, and then only used that detection for display. The conversion, the over-claim guards and the accounting push all read the operator’s manual tick instead. So an invoice whose lines printed inclusive was stored at the inclusive figure, pushed to the accounting system with GST added on top of GST, and the PO over-claim warning fired falsely because it was comparing an inclusive line against an exclusive budget, reporting an overshoot that was exactly the GST. Worse, the manual tick could not disagree with the detection in one direction, because an absent entry was indistinguishable from "exclusive". After. One function is now the single answer that the display, the dropdown, the over-claim guards and the conversion all read. The detection counts as a basis, and the operator’s per-row choice became a genuine three-state override: absent follows the document, and an explicit yes or no wins in either direction. Verified against a live invoice after deploy: the lines now store ex-GST, sum exactly to the invoice’s ex-GST amount, and the pushed bill totals what the invoice printed rather than being overstated by the GST. Impact. This is authored value, not just a display figure. The register writes the claim value that feeds the accounting bill, the subcontract’s certified value accumulation, the retention calculation and the payment schedule. A wrong value there damages the commercial ledger whether or not anyone pushes to accounting, which is why fixing it at source in the register matters more here than anywhere downstream. Manual cash flow cells survive forecast approval Before. Approving a forecast regenerated every cash flow period from scratch and spread it flat across the remaining months. On the linear distribution mode, every manual cell override was silently discarded. On preserve-manual, which was the default, the row was rewritten so every month read as hand-edited, the programme-driven distribution stopped running, and the row degraded to a flat spread with the whole line rendered bold as though you had typed it. After. Manual overrides now survive approval with their amounts intact, the remaining months still follow the programme-task distribution, and only the cell you actually edited renders as edited. Impact. Approving a forecast is meant to lock a number in, not quietly rewrite the shape of it. Every month’s timing on the cash flow depends on this being right. AS 4901 subcontracts print every item you selected Before. Several items in the generated AS 4901 subcontract had print layouts narrower than the form that fed them. The mandatory documents item offered eight checkboxes but the printed manifest only had six rows, so selecting seven or eight silently dropped the overflow. Key personnel and the deed’s subcontractor party details had similar overflow and fill gaps. After. Overflow now condenses into the last row rather than disappearing, key personnel and deed party details fill correctly, and where a field genuinely cannot be auto-filled from upstream data the document explains the gap instead of leaving it silently blank. Party name matching also now works across the punctuation a company name is written with, so "Pty Ltd" and "Pty. Ltd." no longer read as two different parties. Impact. A generated subcontract that silently drops a selection is worse than one that never offered it. What you ticked is now what prints. The welcome email logo is the right shape Before. The Deep Space logo was stretched in the invitation and welcome email. After. It is not. Impact. The first thing a new user sees from us now looks like it was built on purpose. Something here not working the way you expected, or a gap you want closed? Raise it through the feedback portal inside the platform. Every entry above started as one of those.