Product Updates

See the latest features, improvements, and product updates

Powered byCreate yours, for free!
Back to Product Updates

Release notes, 26 Oct to 1 Nov: Today, Team and Data pages, support dashboards, the company learners

Orhan Keskin··12 min read

The busiest week so far: the workspace got a new shape (Today, Team, Data, Settings), support got its own dashboards and a full desk loop, and Argus started to learn each company from its own data.

Highlights

Your workspace

  • Today is the new landing page. This morning lists what needs a person, Waiting for you shows what is yours to decide, and you can ask Jarvis from the same page. (Orhan)
The Today page of a sample workspace: the This morning card with six things that need a person, the Waiting for you card and the Jarvis question box.
  • Team brings departments, agents and their duties into one page. Hire opens a wizard that adds a duty in draft. On a duty's page, Try it on last week makes a test run that sends nothing. (Orhan)
  • Data replaces three sidebar entries with tabs: Connections, Data report, Data catalog, Company profile and Metis. The workspace admin can correct what the Company profile says. (Orhan)
The Data page with the Getting your data in steps and the tabs Connections, Data report, Data catalog, Company profile and Metis.
  • Settings is one page with the tabs General, People, Access, Budget, Notifications, Models, Mail and Privacy. What's new and Help center are in the gear menu. (Orhan)

Dashboards

  • Support and Support board summary are new under Business review, and Support desk under Success. They compare your numbers with your targets and with verified benchmarks, listed under Benchmarks in Settings. (Orhan)
The Support dashboard: tiles for tickets resolved, first response, resolution time, CSAT, NPS and resolved by the agent, with targets, benchmarks and a created versus resolved chart.
  • Support insights and Research are new under Success. Insights reads closed tickets for topic, root cause and help center gaps. Research answers how a question was handled before. (Orhan)
  • Revenue has a redesigned Overview and deeper pages for MRR, Churn, Cash Flow and Past Due. P&L is a new dashboard built on your management accounts. (Sydney)

Slack and connections

  • Slack works both ways. Agent posts carry buttons such as approve and decline, and a reply in the thread goes to the agent. Posts to your own Slack no longer wait for approval. Anything to customers or guests still does. (Orhan)
  • Integrations: Xero connects with one sign-in, and Help Scout is no longer offered. (Orhan)

Everything shipped this week

Agents and automation

  • Team page: departments, agents, duties and hiring are on one page, with a page for each duty, and Engineering is a built-in desk (Orhan)
  • Hire wizard: a few questions add a duty in draft, and a member who sees a department can hire and shape its duties while none is live (Orhan)
  • Duty page: a duty reads as a procedure in one column. You can edit its steps in plain words, tell it what to change, test it on last week without sending anything, and review changes to a live duty in a working copy before publishing them (Orhan)
  • Hand-offs between agents: an agent can pass work to another agent with a deadline, and the Office shows what was handed over and what is late (Orhan)
  • Agent scorecard: each agent gets a weekly score from its results, this week against last, and the general manager gets a weekly ranking by direct message in Slack (Orhan)
  • Department mandates: each department can have a written mandate with history, set in Settings, and every agent run reads it (Orhan)
  • Autonomy guards: an agent that breaks a guard, or scores badly three days in a row, is stepped down one level on its own and never up. The ladder that would raise levels is built, not switched on (Orhan)
  • Shadow week: a duty in draft is compared each night with what the team really did, so its first week can be judged before it goes live (Orhan)
  • Company learners: Argus reads a company's support, marketing, sales, customer, finance, market and Slack history into the Company profile, and re-checks each part on a schedule. Balio runs them, they are not on a schedule of their own (Orhan)
  • Market watch: the market learner finds competitors, reads their public pricing and review pages, counts wins and losses per competitor, and marks every fact with a date after which it is checked again (Orhan)
  • Setup proposal and Setup card: from the Company profile Argus proposes the duties that fit each department, and the general manager can hire them all in draft from one page (Orhan)
  • Fact check before sending: every duty that writes to someone outside has a step that checks the draft against the facts, and the support desk passes it before a reply goes out alone (Orhan)
  • Mailbox assistant: on a person's own connected Gmail it can brief them and draft replies in their tone. Built, every part is off until its owner switches it on (Orhan)
  • Learning from edits: when a person edits or declines a draft, Argus keeps the difference and the reason, asks why in Slack after a decline, and shows the lesson to the agent on similar tickets (Orhan)
  • Working hours and holidays: business hours and public holidays follow the workspace's operating region, and a duty can run every hour at a set minute (Orhan)

Inbox and approvals

  • Today: one landing page for the morning brief, what waits for you and Jarvis (Orhan)
  • Slack both ways: agent posts carry action buttons, and a reply in an agent's thread reaches that agent, which answers there (Orhan)
  • Internal Slack posts: a post to the company's own Slack is sent at every autonomy level. Shared channels, guests and customers keep the approval step (Orhan)
  • Honest counts and notifications: draft counts no longer include items that were never going to be sent, and notifications say what really waits (Orhan)
  • Drafts that go stale: a support draft is replaced when the customer writes again, and old ones are named in the morning brief (Orhan)

Support desk

  • Support loop: tickets move through fixed stages with routing to the right person, bugs go to the engineering tracker, and the customer is told when the fix ships (Orhan)
  • Slack hand-off threads: a ticket that needs a person opens one Slack thread with buttons and reminders, an instruction written there is followed, and a reply from someone who is not a member becomes an Inbox draft (Orhan)
  • Desk actions: the desk can close, snooze and assign conversations on Intercom and Zendesk, each switched on per workspace. Help Scout was removed (Orhan)
  • Auto-send rules: a reply goes out alone only in categories the desk is sure of, the rule is enforced in code, and results of sent replies feed the proposal to widen it (Orhan)
  • Timers and holding messages: the desk follows up, closes quiet conversations and updates bug reporters on a timer, and can send a short holding message while a person decides. Holding messages are switched on per workspace (Orhan)
  • Ticket categories: each workspace has its own list of support categories with an editor and history, and the knowledge base agent is told when a new one is due (Orhan)
  • AnnounceKit's desk: routing, rules and the general manager's decisions of the week were applied, waiting conversations were queued again and stale drafts cleared (Orhan)
  • Research: "how did we handle this before?" searches past tickets, articles and internal documents from Success, Jarvis and Slack (Orhan)
  • Stored threads: conversations that were stored without their messages were filled in, tickets held in a data warehouse now carry their messages, and they count when hiring a support duty (Orhan)

Dashboards

  • Support dashboards: a general manager's view, a support lead's view and a one-page board summary, on one definition of every support number, with targets and verified benchmarks (Orhan)
  • Support insights: one analysis per closed ticket (topic, root cause, help center gap), shown under Success. A fix lets a backlog resume where it stopped (Orhan)
  • Revenue overview: redesigned on the billing figures the data warehouse publishes (Sydney)
  • Revenue deep dives: MRR, Churn, Cash Flow and Past Due pages with concentration, tenure, ageing and growth (Sydney)
  • P&L: a new dashboard on the management accounts. Tiles open a breakdown, and a line that moved between accounts now shows its newest value (Sydney)
  • Filters and periods: Revenue and P&L filter by year, Past Due follows the date range, and Cmd or Ctrl click picks several periods (Sydney)
  • Resellers and individuals: for Fetch, resellers and individual customers are shown beside paying customers (Sydney)
  • Faster dashboards: the slowest pages (Revenue, Customers and the first load of a workspace) were sped up in the database (Orhan)

Data and connections

  • Data page: Connections, Data report, Data catalog, Company profile and Metis are tabs of one page (Orhan)
  • Data report: per source, what arrived, how far back it goes, how fresh it is and what is missing. It refreshes on its own (Orhan)
  • Data catalog: every kind of record Argus holds, where it comes from, how fresh it is and which duties read it (Orhan)
  • Company profile: what Argus learned about the company, each item with its confidence and evidence, and the admin's corrections kept as versions (Orhan)
  • Metis sources and scans: Metis reads Slack history, the help center, the website and a repository's documents into memory, a scan resumes after a stop, and for Slack you choose all public channels or a list (Orhan)
  • One-click connect: a tool can be connected through its own consent screen before a pasted key, and Balio sets its app for each provider inside Argus (Orhan)
  • Xero and other connections: Xero connects by sign-in, Zendesk accepts an access token, a help center kept as files can be read, and testing a Website connection no longer fails (Orhan)
  • Marketing and sales sources: Google Ads, LinkedIn Ads, Apollo sequences and Mailgun tags can be read (Orhan)
  • Warehouse customer identity: one account per customer across billing, CRM and support for workspaces on a data warehouse, and customers an older match had merged were separated (Orhan)
  • Warehouse setup: Argus proposes the mapping that fits a warehouse and a person applies it, tools the warehouse already covers are not asked for, and syncs run after the warehouse's own rebuild (Orhan)
  • Revenue figures from the warehouse: every revenue figure is taken as the data warehouse computes it (Sydney)
  • Complete and fresh data: Argus finds gaps in history and refills them, slows down on sources that have nothing new, marks dead ones, and checks weekly that each tool still sends the fields it expects (Orhan)
  • Data quality score: each workspace has a score from counted defects and freshness, and a duty is held when the data it needs is below the mark (Orhan)
  • Reconciliation across sources: the same monthly figure is compared between billing, the warehouse and finance, and differences are explained (Orhan)
  • Identity links: Argus measures how many billing, support, product and CRM records are tied to an account and proposes the missing links for a person to confirm (Orhan)
  • Logos and photos: accounts and competitors show their logo and members their photo, without calls to outside logo services (Orhan)
  • Unit economics: gross margin, cost per customer and per plan, and a twelve month trend are added to the finance part of the Company profile (Orhan)
  • Business e-mail in the analysis: a person can share business threads from their own mailbox. Only the owner can share, and sensitive mail is dropped before it is stored (Orhan)

Access and security

  • Settings: one page with tabs for General, People, Access, Budget, Notifications, Models, Mail and Privacy, in English (Orhan)
  • Access audit and hardening: an audit of who can read what led to per-person grants for sensitive data, access tiers, suspensions and a team directory, and "See as" never opens an inbox (Orhan)
  • AI spend visibility: spend is shown to the workspace admin and Balio's data team, and closed to members and viewers (Orhan)
  • Retention and deletion: each workspace has a retention policy, a person's data can be erased in one step, and sensitive Slack, calendar and note content is dropped before it is stored (Orhan)
  • Model keys: a workspace can bring its own keys for several AI providers under Models, and Balio's shared key stands in until it has one (Orhan)

Balio's side

  • Evals: agents are tested on old tickets with test sets, several designs and models, a score against cost comparison and a review queue. AnnounceKit's first two back-tests ran (Orhan)
  • Planning by project: the plan is grouped by project (Argus or one workspace) with owners and milestones, and the old setup guide was removed (Orhan)
  • Onboarding a workspace: one command runs connect, report, learn and propose, each run is recorded step by step, and a two-day handbook and the first two pilots are written up (Orhan)
  • Due-diligence report: one page and one file with what Argus holds about a workspace, also on partial data (Orhan)
  • Engine health: a score from 0 to 100 per workspace with checks for silent failures, tasks it opens for itself, and a fallback to another model when one is down (Orhan)
  • Cost control: a cost record per run, cost per result, alerts on unusual spend, a monthly budget split across duties, cheaper models tried where quality holds, and self-repair of stuck work (Orhan)
  • Page views: page opens are recorded, so Balio can see which screens are used (Orhan)

Platform and engineering

  • Visual pass: tables, Settings, the Data report, the Company profile, Metis and the Inbox were redrawn in one visual language, in light and dark (Orhan)
  • What's new and Help center: both open from the gear menu, with an unread count on What's new (Orhan)
  • Faster, safer releases: the test suite runs once per release instead of three times, tests are chosen from what changed, flaky timing tests were fixed, and a new guard stops a database change from undoing an earlier one (Orhan)
  • Diagnose tools: dozens of read-only checks were added for support, evals, the warehouse, access and the database, so a question about production is answered from data (Orhan)