
GPT-6 Astra vs Sol vs Luna: which AI employee fits sales?
Compare Astra, Sol and Luna on 20 fictional catering enquiries: completion time, lead ratings, money rules and drafts a reviewer would send unchanged.
KyoubeAI · · 7 min read
Use cases
A pipeline that works itself while your people talk to customers

This suite is built from parts KyoubeAI runs today: AI employees, data tables, apps, connections and routines. Explore its applications and workflows, then talk to us about scope, connected systems and licensing for your team.
Every sales team has the same two problems. The CRM is out of date because updating it is nobody's favourite job, so the forecast is a guess. And follow-up is inconsistent, so deals die in silence rather than being lost for a reason anyone can learn from.
The Sales Supercube gives the team a working sales operation in which the agents do the updating and the following-up. It seeds tables for accounts, contacts, leads, deals, activities, quotes and forecasts. It declares agents that qualify inbound leads, keep every deal record current from email and calendar, draft quotes from the price book, chase the next step, and roll up the forecast with the reasoning shown. It publishes apps for the pipeline board, the account page, the quote builder and the forecast.
If you already run HubSpot, Pipedrive or Salesforce, the supercube keeps that CRM as the system of record and mirrors what it needs. If you do not, the seeded tables are the CRM, and they are yours.
Once it is running, a lead fills in the website form. The SDR agent enriches it, scores it against your ideal customer profile, writes a two-line summary and books it into the right person's calendar if it qualifies. After each call the Deal Desk agent reads the calendar and the email thread, updates the stage, next step and close date, and drafts the follow-up for the salesperson to send. On Friday the Forecaster agent has the commit, best case and pipeline numbers with the deals that moved them, and the sales manager spends the pipeline review on the deals, not on the spreadsheet.
Companies with two to fifty people selling: founders who sell, small B2B teams, wholesale and distribution reps, agencies with a business-development lead. The roles that use it are sales reps, sales development reps, sales managers and the founder or MD who wants a forecast.
| Agent | Role | Owns | Harness preference |
|---|---|---|---|
| SDR | general | Inbound leads: enrichment, scoring, qualification, routing, first-touch sequences | Claude Code, then pi |
| Deal Desk | general | Deal hygiene: stages, next steps, close dates, activity capture from email and calendar | Claude Code |
| Quote Assistant | general | Quotes and proposals from the price book; discount checks; e-signature routing | Claude Code |
| Account Manager | general | Existing customers: renewals, expansion signals, health, check-in cadence | Claude Code, then Hermes |
| Forecaster | analyst | Weekly forecast, win/loss analysis, pipeline coverage, rep scorecards | Claude Code |

| Table | Purpose | Key fields |
|---|---|---|
accounts | Companies you sell to, shared with Finance as customers | name, industry, size, owner, tier, crm_id |
contacts | People at those accounts | account, name, role, email, phone, consent |
leads | Inbound and outbound leads before they become deals | source, contact, score, icp_fit, status, routed_to |
deals | Opportunities through the pipeline | account, title, stage, value, close_date, next_step, owner, probability |
activities | Calls, emails, meetings and notes against deals and contacts | deal, contact, type, at, summary, by |
quotes | Quotes and proposals with lines and status | deal, version, lines, total, discount, status, sent_at, signed_at |
price_book | Products, services, list prices and discount rules | sku, name, unit, price, floor_price, approver_above |
forecasts | Weekly forecast snapshots per rep and team | week, owner, commit, best_case, pipeline, closed |
win_loss | Reasons recorded when a deal closes | deal, outcome, reason, competitor, learned |
sequences | Follow-up sequences and their steps | name, steps, audience |
| Group | Target systems | Flows in | Flows out |
|---|---|---|---|
| CRM | HubSpot, Pipedrive, Salesforce | Accounts, contacts, deals, activities | Stage, next step, close date, notes, new contacts |
| Email and calendar | Gmail, Outlook, Google Calendar, Microsoft 365 | Threads and events on connected mailboxes | Drafted follow-ups, meeting invitations |
| Documents and signature | DocuSign, PandaDoc, Google Docs | Signed status | Quotes and proposals |
| Prospecting | LinkedIn Sales Navigator (exports), enrichment APIs | Firmographics, contact data | Nothing |
| Chat | Slack, Microsoft Teams | Deal updates posted by reps | Won-deal announcements, stale-deal nudges |
| Finance | Xero, QuickBooks, via the Finance Supercube | Invoice and payment status per account | Won deals as draft invoices or contracts |
Inbound lead to first meeting Trigger: a form submission or a lead handed over by Marketing. Steps: the SDR enriches the company and contact, scores fit against the ideal customer profile, checks for an existing account, writes the summary, and either routes to a rep and offers meeting slots, or starts a nurture sequence, or disqualifies with a reason. Checkpoint: none for routing; a rep accepts or reassigns in the lead inbox. Output: lead qualified, meeting booked, record created in the CRM.
Deal hygiene after every meeting Trigger: a calendar event with a deal contact ends. Steps: the Deal Desk reads the event and the thread, summarises the meeting as an activity, proposes the stage, next step and close date, and drafts the follow-up email. Checkpoint: the rep confirms the stage change and sends the email. Output: the deal is current and the follow-up is out within the hour.
Stale deal chase Trigger: routine, daily 08:30. Steps: for each open deal with no activity for longer than its stage allows, the Deal Desk nudges the rep with a suggested action; after a second period it flags the deal to the manager and proposes closing it as lost with a reason. Checkpoint: the rep or manager decides. Output: pipeline stays honest.
Quote to signature Trigger: a rep asks for a quote on a deal. Steps: the Quote Assistant builds the quote from the price book, checks every discount against policy, generates the document and routes for approval if a discount exceeds the rep's authority. Checkpoint: discount approval by the manager where required; the rep sends. Output: quote sent for e-signature, signature status tracked, won deal handed to Finance.
Weekly forecast Trigger: routine, Friday 12:00. Steps: the Forecaster snapshots every deal, categorises commit and best case from stage, probability, age and activity, compares to last week, and writes the changes. Checkpoint: each rep confirms their commit; the manager approves the team number. Output: forecast published to the forecast app and the General Management scorecard.
Renewal and expansion Trigger: routine, daily, 90 and 60 and 30 days before a contract end date. Steps: the Account Manager checks usage or support signals, invoice status from Finance and recent activity, drafts a health note and a check-in email, and opens a renewal deal at 90 days. Checkpoint: the account owner sends the check-in. Output: no renewal is a surprise.
B2B services agency. Deals are proposals with scopes rather than SKUs; the Quote Assistant builds proposals from a scope library with day rates. Win/loss reasons feed the Marketing Supercube's messaging. Forecast coverage is measured against capacity from the Operations Supercube.
SaaS company. Leads are product sign-ups as well as forms; the SDR reads product usage to score them. Deals carry ARR and term; the Account Manager watches usage drops as churn signals. Won deals create the subscription record that Finance recognises revenue from.
Wholesale distributor. Accounts are retailers with reorder patterns; the Account Manager predicts reorder dates from the Supply Chain Supercube's order history and drafts the call list. Quotes are price lists by customer tier with volume breaks.
Manufacturer. Long deals with engineering involvement; the Deal Desk tracks sample and specification milestones as stages. Quotes reference bill-of-material costs from Supply Chain so margin is checked before a discount is approved.
Estate agency. Accounts are vendors and landlords; deals are instructions and lets; contacts are buyers and applicants matched to properties held in an Operations asset table. The SDR handles portal enquiries and books viewings.
Recruitment agency. Two pipelines, clients and candidates, sharing the same tables with a kind
field. The Deal Desk tracks placements from brief to start date; the Forecaster predicts fee revenue from
interview-stage counts.
write on its own tables and read on Finance's invoice tables. CRM writes go
through the connector under a tool profile that lists the exact operations allowed.Step-by-step guides that set up part of this work on KyoubeAI.

Compare Astra, Sol and Luna on 20 fictional catering enquiries: completion time, lead ratings, money rules and drafts a reviewer would send unchanged.
KyoubeAI · · 7 min read

Use AI employees to organise inbound sales enquiries, draft relevant follow-ups and keep each lead moving through a shared pipeline with human review.
KyoubeAI · · 7 min read
Suites share tables and compose, so the ones next to this one are the ones it works with.

The operating rhythm of the company, run by agents, owned by the leadership team
Goals, KPIs, meeting cadence and decisions in one place, with agents that prepare every review and chase every action so leaders spend their time deciding, not collecting.

Books that close themselves, cash you can see, approvals that stay with people
Accounts payable, receivable, month-end close, cash forecasting and management reporting run by agents against your own ledger data, with every payment and every journal approved by a person.

Campaigns planned, produced, published and measured from one place
Content, campaigns, channels and attribution run by agents against your own marketing data, with brand rules enforced before anything is published and every result tied back to a lead.