# GenieWorks website copy Generated from the rendered homepage, product detail pages, and stories.js. Run export-copy.mjs to refresh. # Homepage ## The promise ### Like the best engineer you’ve worked with Your AI as an Engineer inside Jira, GitLab, or GitHub GenieWorks gives your AI an engineer’s mindset. It challenges requests that won’t solve the real problem, recommends a better approach, and checks the result against what your team agreed. Jira Product / APP-42 ### Stop duplicate exports GenieWorks Specification Activity Comments & history - Alex Developer 9:41 Assigned to GenieWorks Specification started · Investigating export retries - GenieWorks AI engineer 9:42 Pulled live evidence via MCP Sentry One click, a lost response, then two export jobs. Internal API Same user and export settings; the retry has a new request ID. - GenieWorks AI engineer 9:44 Needs input · During specification A button lock won’t cover network retries. Should a retry return the original export? - Alex Developer 9:45 Yes, reuse the same export. @Sam, please confirm from QA. - Sam QA 9:46 Agreed. Only “New export” creates another. Test retries after a lost response. - GenieWorks AI engineer 9:48 Specification updated · Ready for review Retries reuse the export; “New export” creates another QA’s lost-response scenario added to the test plan Review changes Original request Disable Export after the first click. Live evidence · Sentry + your Internal API Sentry traces link one click to a lost response and two jobs. The internal service confirms a new request ID on retry for the same user and export settings. These authorized MCP reads supply context the repository alone cannot show. Agreed behavior Keep the same export across clicks and network retries. Create another only through “New export”. QA verification added Simulate a lost response, retry, and check that the original export is returned. Explore specification Implementation - Alex Developer 10:03 Assigned to GenieWorks Spec approved · Implementation started - GenieWorks AI engineer 10:20 PR #84 opened · Implementation complete Acceptance criteria verified; configured checks passed Found and fixed a Web retry gap during AC review Review implementation Gap found and fixed The API reused the export, but the Web retry generated a new request. Corrected before opening the PR. Verified against the agreement Lost-response retries reuse the export. An explicit new export creates another. Review scope This example has AC review enabled. Human review and release approval remain separate. - Alex Developer 10:24 Please also link the retry message to the existing export. - GenieWorks AI engineer 10:29 PR #84 updated Retry message now links to the original export AC updated; checks and acceptance review repeated Acceptance review enabled Explore implementation Code assurance - Sam QA 11:02 Assigned to GenieWorks Code assurance started - GenieWorks AI engineer 11:12 9 scenarios reviewed · 1 supported finding The planned retry cases show no gap. The problem appears when cleanup overlaps a long-running export. - GenieWorks AI engineer 11:14 Retry record removed while export is still running Cleanup can expire the retry record before the export finishes. A late retry then has a path to create a second job. Finding survived independent review Source evidence and proposed verification attached Review 9 scenarios 8 scenarios without a finding in source review First export; repeated click; lost-response retry; retry after completion; explicit new export; concurrent retries; permission denial; worker restart. Scenario 9 · Long-running export + cleanup + late retry Cleanup expires retry records by age while the worker can still be processing the linked export. The agreed cases did not cover that overlap. Proposed verification Keep an export running beyond retry-record retention, run cleanup, then retry. Verify it returns the existing job. These steps have not been executed. 9 scenarios analyzed in source · Application, API, and browser tests have not been run in this example. Explore code assurance APP-73 · Propose customer self-service onboarding Project proposal - Maya Product lead 14:00 Assigned to GenieWorks · Ran “Create Project Proposal” Saved team recipe · Inputs: self-service onboarding, enterprise customers - GenieWorks AI engineer 14:08 We can reuse account provisioning. Public signup adds billing and abuse controls. Start with an invited enterprise pilot? - Maya Product lead 14:11 Yes, pilot first. @Alex, can we keep our current account approval? - Alex Developer 14:14 Yes. Reuse provisioning and manual approval. Add guided setup after the invitation. - GenieWorks AI engineer 14:20 Project proposal ready for review Pilot scope, delivery phases, risks, and success criteria Team decisions recorded in the proposal Review proposal Scope · Invited enterprise pilot Reuse provisioning and account approval. Build guided setup and invitations. Public signup, billing changes, and abuse controls stay outside the pilot. Delivery phases Validate the onboarding journey, deliver setup across Web and API, then onboard the first invited customers. Success and risks Customers finish setup without developer intervention. Measure completion and support requests; check account permissions before pilot launch. Decision still needed Product selects pilot customers and launch timing. The proposal is ready for review, not approved for delivery. Your recipe defines the inputs, investigation, and deliverable. The team can run it again on the next proposal. Explore custom workflows Project workspace Lessons & decisions ### What I learned from our last sprint GenieWorks & project leads · Shared learning review Project learning - GenieWorks AI engineer 9:00 I reviewed 18 tickets I worked on, bugs filed afterward, and changes to Jira and Git after my implementation. - GenieWorks AI engineer 9:04 I missed a lifecycle case APP-31 led to bug APP-57. Alex’s corrective PR #79 keeps retry records until the export finishes. My spec covered retries, but missed cleanup while a job was still running. - Alex Tech lead 9:10 Yes. Apply that lesson to background jobs with retries, not every API endpoint. - Maya Product lead 9:12 Agreed. Keep “New export” as an explicit new action. - GenieWorks AI engineer Next matching task Lesson checked, adopted, and applied Replayed APP-31: the missing cleanup scenario now appears APP-86: added the lifecycle check without another reminder See what changed Evidence behind the lesson Compared APP-31’s original spec and PR #62 with later bug APP-57, the Jira correction, and Alex’s corrective PR #79. The changed behavior belonged to the original scope; this was not a new feature or release cherry-pick. What I now check For retryable background jobs, specify what happens when cleanup overlaps a running job and a late retry. Keep request identity until the job finishes; preserve an explicit new action. Checked before reuse Replayed the original ticket: the proposed lesson adds the missing acceptance and test scenario. Checked a read-only request too: this lesson does not apply. Applied on APP-86 The next export specification includes the cleanup, running-job, and late-retry scenario, with a link to this approved lesson. Future reviews check whether the lesson was useful or needs refining. Lessons checked against outcomes, reviewed with leads, and reused where they fit. Explore knowledge & learning The work, questions, and decisions stay on the issue See how the work moves forward Your model · Your infrastructure · Shared team context Works with you in - Jira - GitLab - GitHub Issues Works with your model - Codex - Claude - Qwen Where your AI runs - Hosted - Your environment Explore deployment options ## Proptech SaaS Case Study ### Why an Engineer’s mindset matters A strong, fast-moving Proptech SaaS team already used AI in daily development. GenieWorks brought that capability into a shared engineering process: investigated requests, clear agreements, and implementation reviewed against the intended outcome. Project context and decisions carried through the work, giving the whole team a common way to deliver with AI. Read the case study and how it was estimated ### Proptech SaaS Case Study About 4× estimated effective delivery capacity Same team size · Already using AI · March vs January Estimated delivery capacity Jan–Jun 2026 1.0×Jan 1.9×Feb 3.9×Mar 7.6×Apr 8.0×May 7.3×Jun January = 1× · Select a month to explore ## Know where the AI budget goes ### More tokens ≠ more value A stronger model can build the wrong solution exceptionally well. GenieWorks investigates the evidence, questions assumptions, and brings doubts to your team. When a decision needs your input, it asks. Together, you clarify what should change and why. That agreement guides implementation and review. See where the effort went, from overall spend to the ticket, model calls, and recorded activity. SpendTicketModel callActivity The spending overview ### Follow the work Behind the spend Open a ticket to see who started the work, which model ran, and how usage breaks down across investigation, specification, and review. Inside the example APP-58 used $64.73. About 74% of its tokens went into investigation. The activity shows the same API schema read three times. That gives your team a specific pattern to investigate, improve, and compare on the next run. Explore cost reporting and run traces The ticket, the calls, the repeated reads ## Delivery your team can review ### An engineer who follows through With your team in the loop GenieWorks keeps questions, evidence, and decisions on the issue, where teammates can contribute and understand why an approach was chosen. Those agreements guide implementation and review. Progress, checks, and open questions stay visible. Corrections become reviewed lessons for future work. Explore implementation Explore shared knowledge - Decisions the whole team can followSee the evidence, questions, and agreements behind the chosen approach. - Visible follow-throughSee what changed, what was checked, and what still needs attention. - Lessons applied to future workCorrections become reviewed lessons that shape how similar tasks are handled. ## Pricing PRICING ### One complete suite The same full suite at every team size. Pricing follows your Jira site’s user count. | What’s included | 1–10 users $60USDtotal per Jira site per month Contact us | 11+ users $13USDper Jira user per month Contact us | Enterprise CustomBy agreement Let’s Talk | | --- | --- | --- | --- | | Project knowledge | Included | Included | Included | | Specification | Included | Included | Included | | Implementation & review | Included | Included | Included | | Code assurance | Included | Included | Included | | Usage & cost reports | Included | Included | Included | | Priority support | — | — | Included | | SLA | — | — | Included | | Single sign-on (SSO) | — | — | Included | | Custom models | — | — | Included | | On-premise installation | — | — | Included | View monthly totals by Jira site size **Full suite · Total USD per Jira site per month** | Licensed Jira users | Total USD / month | | --- | --- | | 1–10 | $60 | | 11–15 | $195 | | 16–25 | $325 | | 26–50 | $650 | | 51–100 | $1,300 | | 101–200 | $2,600 | | 201–300 | $3,900 | | 301+ | Let’s Talk | Jira pricing · USD. Full-suite totals for the whole Jira site at each tier’s maximum user count. The standard $13/user rate applies above 10 users. Model-provider charges are separate. All licensed Jira site users count. Jira pricing · USD. For GitLab or GitHub, talk to us . On-premise execution: AI runs in your environment; coordination stays hosted. Deployment details ## Bring shared AI into your delivery process ### Make AI part of how your team delivers For teams evolving an existing product, where shared context, review, and a clear definition of done matter. Let’s Talk Explore the engineering process # Specification — GenieWorks Page: [Agree on the right change before implementation begins](./specification.html) ## Specification — from request to agreement SPECIFICATION ### Agree on the right change before implementation begins GenieWorks investigates the request, brings the decisions to your team, and turns the agreed outcome into work an engineer can build and verify. Follow one request ### “Disable Export so customers stop getting duplicates” THE AGREEMENT WE REACH A retry returns the original export. An explicit New export creates another. Follow the decisions below Email, Slack, or Zendesk → your existing Jira intake → GenieWorks ## Investigate the cause and bound the change 01 · INVESTIGATE ### One click can still create two exports GenieWorks checks the reported behavior against the code and live evidence before accepting the proposed fix. Here, Sentry records a lost response. The connected Internal API shows a second job after the retry. Disabling the button would leave that path untouched. The evidence changes the fix - 1The first request creates a jobThe response never reaches the browser - 2Web retries the exportA new request identity reaches the API - 3The API accepts another jobThe same intent has become two exports Selected project MCP connections 02 · BOUND THE CHANGE ### Follow the whole path. Change only what the fix needs. A queue redesign would add work without resolving the missing agreement about request identity. This example needs coordinated Web and API changes. ### Preserve the request Keep its identity after a lost response. Give a deliberate new export a fresh identity. ### Reuse the authorized export Recognize the retry and return the existing job within the caller’s allowed scope. ### Keep file generation unchanged The existing worker can process the single accepted job. No worker task is needed for this fix. Other investigations can use Prometheus, Grafana, Better Stack, or your own internal services through compatible MCP connections selected for the project. ## Resolve the product choice together 03 · AGREE ON THE BEHAVIOR ### Ask the question that changes the product The code can explain why duplicates happen. It cannot decide what your customer should see when they retry. GenieWorks raises that choice with its evidence. The answer and QA’s agreement stay with the issue, ready for implementation to use. APP-42 · Comments & history - GenieWorks AI engineerThe retry can reuse the job. Should it return the original export or show a duplicate warning? - AAlex DeveloperReturn the original export. @Sam, does that match the behavior QA expects? - SSam QA leadAgreed. Only an explicit New export should create another job. Include denied access in the checks. - GenieWorks AI engineerUpdated the specification: retry reuse, explicit New export, and permission checks now have separate acceptance criteria. Team decision carried into the spec ## Inspect the agreement before Apply 04 · MAKE IT EXECUTABLE ### The next engineer inherits the decisions The specification connects the evidence and agreed behavior to component changes, acceptance criteria, and planned checks. Open the decision and the resulting specification below. SPECIFICATION REVIEW ### “Handle retries” still leaves too much to guess A separate item-level review finds the vague instruction and makes the required behavior explicit before handoff. Before review Handle retries safely After correction Web keeps the original request identity after a lost response. The API returns the existing authorized export. New export creates a fresh identity. Automatic specification correction enabled in this example. Default review records gaps; correction is optional. The team reviews and chooses Apply. The test plan describes checks to run; it is not a test result. Applying a specification does not approve code or release a change. Explore the specification walkthrough ## Apply the same discipline to other requests BEYOND THIS BUG ### Different requests need different questions The same shared project context supports a product idea or a supplied design. What changes is the investigation and the agreement the team needs. ### A rough product idea “Let customers schedule exports” leaves frequency, permissions, failure handling, and ownership undecided. GenieWorks works through those decisions and the affected components before turning the idea into scope, acceptance criteria, and a test plan. A SUPPLIED DESIGN · SEPARATE EXAMPLE ### Carry the design into the requirements An approved export-history drawer should not become a convenient modal. GenieWorks maps the design to your existing UI and specifies what to reuse, adapt, or create. Placement, action order, selected-row visibility, responsive behavior, and loading, empty, and error states become requirements. Unresolved differences become questions. Design reference + existing UI → visual requirements Design handoff · Export history Approved design Export history 1 All exports Last 7 days | Export | Status | | --- | --- | | September | Ready | | August | Ready | | July | Ready | 1–3 of 12 · Page 1 2 Export details September export 3 Retry history - Request received - Retry · same job - Export ready Download · Close - 1 Reuse Existing table, filters, and pagination. Keep grouping, row density, and column order. - 2 Adapt Existing details drawer. Add retry history; keep right-hand placement and action order. - 3 Create The approved design needs a retry timeline. In this example, no existing pattern fits. Must stay true Filters above the table · Details in the right-hand drawer · Selected row stays visible Mobile layout and loading, empty, and error states come from supplied requirements or existing patterns. Unresolved differences stay as questions. ## Make the engineering approach a team capability THE APPROACH TRAVELS WITH THE TASK ### Your team sets the choices GenieWorks selects the approach GenieWorks assesses complexity and selects a specification profile. Your team controls the available models, reasoning depth, instructions, and connected tools. The settings below show those choices. - Triage simple taskLuna · Medium reasoningA bounded fix with focused evidence and checks - Triage complicated taskSol · High reasoningRoot cause, component impact, and product decisions - Create design specSol · High reasoningUser journeys, existing UI, and visual requirements ### A clear agreement makes the model handoff possible In our workflow, GPT-5.6 Sol at high reasoning plans the work; Luna implements against the specification and the same acceptance criteria. GenieWorks recommends an implementation model and reasoning level from the agreement and your configured choices. 100× lower implementation model price Owner-reported Sol-to-Luna price comparison. Planning still uses Sol. This is not a total-task saving or a claim about fewer tokens. Our hands-on estimate: hundreds of developer hours saved on investigation and specification. Experience-based estimate, not a measured customer result. ## Carry the specification into implementation THE AGREEMENT IS THE STARTING POINT ### Now carry it through the code Follow APP-42 into implementation, where the same criteria expose a missed retry path. Follow the implementation # Implementation — GenieWorks Page: [Build the agreed outcomeCheck the work against it](./implementation.html) ## Implementation — carry the agreement through the code IMPLEMENTATION ### Build the agreed outcome Check the work against it GenieWorks carries the accepted specification through coding, verification, and acceptance review. Your team receives a PR with an account of what changed and how it was checked. Follow the implementation ### A retry returns the original export Alex assigns the agreed Web/API work THE RESULT WE’LL INSPECT PR #84 · Retry gap corrected Configured checks and AC review repeated AC review enabled Read the specification this work starts from ## Build and review against the accepted scope 01 · BUILD WITH A SHARED TARGET ### The agreement stays beside the code The accepted handoff defines Web retry identity, API reuse, and permission behavior. GenieWorks implements the selected component work in isolated execution and runs the project’s configured checks. Then acceptance review checks the committed change against the criteria and implementation checklist. In this example, passing checks still leave one agreed behavior missing. AC-01 · THE BEHAVIOR TO DELIVER ### A lost-response retry returns the original export It must create no second job. AC review enabled 02 · FIND THE GAP AND FOLLOW THROUGH ### The API reuses the export Web still starts another request The API change is locally correct, but Web creates a fresh request identity after a lost response. Together, they still produce a duplicate. Review ties that gap directly to AC-01. GenieWorks corrects the Web retry, adds its regression check, and repeats affected verification and acceptance review. A finding that remains unresolved stays in the report. Original request req-42 Web retry req-43 Original request req-42 Web retry req-42 New export remains a deliberate action with a fresh identity ## Report the result and incorporate the team’s follow-up 03 · REPORT THE WORK ### The team can see what was delivered GenieWorks opens PR #84 and reports the implemented behavior, the gap found and corrected, and the checks performed. Alex asks for one addition. That changes the agreement, so the update is implemented and reviewed against the revised criteria. APP-42 · Comments & history - AAlex DeveloperAssigned to GenieWorks · Accepted specification ready for implementation - GenieWorks AI engineerPR #84 opened. Web/API retry behavior implemented. AC review found and fixed the Web identity gap. Configured checks and acceptance review repeated. - AAlex DeveloperPlease also link the retry message to the existing export. - GenieWorks AI engineerPR #84 updated. The message links to the original export. AC-01 updated; checks and acceptance review repeated. ## Inspect the implementation handoff 04 · INSPECT THE HANDOFF ### A PR with evidence behind “done” Inspect the agreement, the missed obligation, and the final report. The final handoff names what changed, what was checked, and what still needs human review. Human review and release remain separate. The report covers the agreed scope and configured checks. It is not a claim of exhaustive testing or production validation. Explore the implementation walkthrough ## Make follow-through part of the team workflow FROM ONE TASK TO THE NEXT ### The workflow carries the discipline The accepted specification stays the target across coding, review, and repair. Project conventions and configured checks travel with the work. Teammates can inspect the decisions and evidence in the shared issue. Feature tags: Agreed scope · AC-based review · Repair & recheck · Visible handoff. ## Explore behavior beyond the agreed checks WHEN YOU WANT A DEEPER LOOK ### What about the scenarios the agreement didn’t cover? Code assurance explores broader interactions and challenges the findings. It is a separate option for deeper investigation. Explore code assurance # Code assurance — GenieWorks Page: [Examine the behavior beyond the planned checks](./code-assurance.html) ## Code assurance CODE ASSURANCE ### Examine the behavior beyond the planned checks GenieWorks explores scenarios, traces reachable source paths, and challenges candidate findings. Your team gets a supported explanation of what may fail and how to verify it. Follow the work ### What if cleanup runs before the export finishes? WHAT THE TEAM RECEIVES 9 source scenarios · 1 supported finding A concrete verification scenario to run ## The ordinary retry works The lifecycle overlap is different 01 · EXPAND THE SCENARIOS ### The ordinary retry works The lifecycle overlap is different Sam assigns the implemented export behavior for code assurance. The agreed first-request, retry, and permission cases do not reveal a gap. GenieWorks follows the related lifecycles. A long-running export can outlive the record that lets the API recognize its retries. Cleanup and a late retry now matter together. - 1The first export is still runningFile generation has not completed - 2Cleanup removes its retry recordThe record expires independently of the job - 3The customer retries the original requestThe API no longer finds the record and can accept another job ## A plausible concern has to survive review 02 · CHALLENGE THE FINDING ### A plausible concern has to survive review The investigation derives nine scenarios and traces them through source. One candidate links cleanup eligibility to retry lookup and new-job creation. A separate review asks whether that path is reachable, whether another guard prevents it, and whether the expected behavior is supported. The finding stays only if its evidence survives those questions. Eight scenarios have no supported finding. That does not mean eight application tests passed. APP-42 · Code assurance history - SSam QA leadAssigned to GenieWorks · Investigate scenario coverage after implementation - GenieWorks AI engineer9 scenarios analyzed. 1 supported finding. Cleanup can remove retry protection while the export is running. A late retry can then reach new-job creation. - GenieWorks AI engineerIndependent source review supports this path. The report includes the setup, source reasoning, and proposed verification. Application execution remains to be done. ## Inspect the finding and its limits 03 · INSPECT THE RESULT ### A finding the team can investigate Open the supported finding and its scenario. The report separates expected behavior, the reachable failure, source evidence, and what remains uncertain. Inspect the nine source scenarios - S-01First authorized requestNo supported finding - S-02Lost-response retryNo supported finding - S-03Explicit New exportNo supported finding - S-04Denied accessNo supported finding - S-05Concurrent retriesNo supported finding - S-06Worker failure and retryNo supported finding - S-07Worker restartNo supported finding - S-08Cleanup after completionNo supported finding - S-09Cleanup during a running exportSupported finding · F-01 Source-analysis coverage. None of these rows claims an executed application test. Source analysis, not executed application tests. The proposed check holds the original worker, advances the retry-record age, runs cleanup, and retries the same request. Browser, API, and production behavior have not been observed in this example. ## Make the investigation repeatable A SHARED TEAM CAPABILITY ### Make the investigation repeatable Scenario reasoning, explicit expectations, and independent challenge become part of the workflow. The team receives the evidence needed to decide on a corrective task and a lesson for future work. Feature tags: Scenario coverage · Challenged findings · Source evidence · Explicit limits. ## Carry the finding into the next task CONTINUE THE WORK ### Carry the finding into the next task Correct the supported gap, then examine what the team should learn from it. Return to implementation See how the learning works # Project knowledge and learning — GenieWorks Page: [An engineer who learns your systemAnd learns from the work](./knowledge.html) ## Knowledge & learning KNOWLEDGE & LEARNING ### An engineer who learns your system And learns from the work GenieWorks builds shared project understanding, examines later bugs and rework, and proposes lessons. Your team can inspect what it learned and how the next task changes. Follow the work ### A later bug reveals a missed lifecycle case APP-31 → APP-57 / PR #79 → APP-86 WHAT THE TEAM RECEIVES A scoped, reviewed lesson The missing check appears in the next specification ## Establish shared project understanding 01 · UNDERSTAND THE SYSTEM ### Give each task the same foundation Project Scan investigates the source and delivery boundaries. Project Knowledge exposes the resulting architecture, component ownership, and engineering rules for your team to inspect. ### Own the user’s intent Keep request identity through a retry; distinguish a deliberate new export. ### Know the boundaries Separate request acceptance, authorization, and asynchronous file generation. ### Keep acceptance explicit Name the required behavior and the checks that will demonstrate it. A validated initial Knowledge baseline can enter use awaiting review. Later replacements require an authorized activation decision. Each run records the context revision it used. Explore how project understanding is built ## The delivered work teaches something new 02 · EXAMINE WHAT HAPPENED LATER ### The delivered work teaches something new In this earlier example, GenieWorks compares APP-31 with bug APP-57 and corrective PR #79. The original specification handled ordinary retries; the later correction exposed the missing cleanup/running-job overlap. It separates an obligation the work missed from a newly requested feature, then proposes the reusable lesson. The lead narrows its scope before activation. Project learning · Conversation with leads - GenieWorks AI engineerI linked APP-31 to APP-57 and PR #79. We missed what happens when cleanup removes retry protection before the export finishes. I propose adding that lifecycle investigation to matching work. - AAlex Engineering leadKeep the lesson specific to asynchronous jobs with expiring retry records. Do not add it to every API change. - SSam QA leadCheck it against the original task and an unrelated UI task. We need the missing scenario without unrelated requirements. - GenieWorks AI engineerThe replay adds the missed lifecycle case to the original specification and leaves the UI task unchanged. The scoped lesson is ready for review. ## Demonstrate the learning on a later task 03 · REVIEW, ACTIVATE, AND REUSE ### Show what changed without another reminder After the lesson is reviewed and active, APP-86 includes the lifecycle criterion and planned check. Its run records the lesson used. This is a subsequent example, not a claim that APP-42 produced this history. APP-86 · THE VISIBLE DIFFERENCE ### The specification now tests the overlap The team does not have to reconstruct the correction in another prompt. Earlier agreement A retry returns the original export Later matching task A retry still returns that export when cleanup becomes eligible while the job is running. Hold the worker and exercise that overlap in the test plan. The replay demonstrates changed requirements and planned verification. It does not claim that the application scenario has already been executed. ## Keep learning scoped and inspectable A SHARED TEAM CAPABILITY ### The next task starts with what the team learned Corrections and later delivery evidence can improve the working method. Your team reviews the proposed lesson, sees where it applies, and can inspect the revision selected for a run. An export-lifecycle lesson leaves unrelated work alone. Feature tags: Initiated learning · Lead feedback · Replay evidence · Visible reuse. ## Put that understanding to work CONTINUE THE WORK ### Put that understanding to work See how shared context becomes evidence, scope, and acceptance in a new specification. Explore specification # AI cost reporting — GenieWorks Page: [Follow AI spendingto the work behind it](./reporting.html) ## AI cost reporting AI COST REPORTING ### Follow AI spending to the work behind it GenieWorks connects usage to the task, person, model, and outcome. Follow an expensive run into its calls and activity, then decide what to improve. Follow the work ### About $10,132 a month What produced the spend? $60,792.85 over six months · 1,248 runs WHAT THE TEAM RECEIVES One ticket → its calls → repeated reads A specific improvement to evaluate Estimated API-equivalent cost, not an invoice ## Start with who, when, and what ran 01 · FIND THE WORK BEHIND THE TOTAL ### Start with who, when, and what ran The monthly view shows totals and changes over time. The ticket list makes the spending concrete: which task ran, who started it, which model it used, and its estimated cost. Here, Maya’s APP-58 specification run stands out. Follow that run before deciding whether the answer is a different model, a different method, or justified investigation. APP-58 · SLOW EXPORTS ### Maya · June 18 · Claude ## Three full reads deserve a closer look 02 · INSPECT THE INVESTIGATION ### Three full reads deserve a closer look APP-58 used 9.52M tokens in repository investigation, 2.406M in specification, and 1.02M in acceptance review. Inside the investigation, api-schema.json was read in full three times. A strong prompt can repeat this expensive discovery when the repository does not expose a useful, smaller contract surface. The trace gives you a cause to investigate, not just a larger total. - 109:14:08 · Read api-schema.json738 KiB returned - 209:16:42 · Read the same file738 KiB returned - 309:19:11 · Read the same file again738 KiB returned ## Inspect one continuous cost investigation 03 · FOLLOW THE EVIDENCE ### From the monthly picture to the next experiment The gallery keeps one dataset throughout: overview, APP-58, and its improvement proposal. Ticket rows are a subset of the period total; file bytes do not establish exact per-file tokens or cost. ## Make the next comparable run more informative 04 · TEST THE IMPROVEMENT ### Make the next comparable run more informative The proposal is to select the relevant schema objects before reading the API contract. A line limit alone would not bound a single-line JSON file. Try that change on equivalent work with the same model and scope. Compare input tokens and total usage, then check that the resulting specification still covers the required behavior. Savings have not been measured in this example. Lower usage alone does not establish a better result. THE COMPARISON TO RUN ### Same task and model A narrower investigation - Record the repository and workflow revisions - Compare investigation input and total usage - Review acceptance coverage and output quality - Keep the change only if the evidence supports it ## Keep the explanation with the work A SHARED TEAM CAPABILITY ### Keep the explanation with the work The next teammate can inspect the initiator, run, calls, and evidence without reconstructing someone’s private session. The improvement becomes a recorded comparison the team can repeat. Feature tags: Task attribution · Model-call usage · File activity · Quality-aware comparison. ## Choose the model with the work in view CONTINUE THE WORK ### Choose the model with the work in view Inspect model and deployment choices, or see how a clear specification enables a cheaper implementation model. See the model handoff Models & deployment # Custom workflows — GenieWorks Page: [Give your workflowa Run button](./recipes.html) ## Custom workflows CUSTOM WORKFLOWS ### Give your workflow a Run button GenieWorks turns your team’s method into a shared tool. Choose a recipe, supply its business inputs, and let the workflow carry the investigation, decisions, and result. Follow the work ### “Help enterprise customers onboard themselves” Maya · Product lead · Saved team recipe WHAT THE TEAM RECEIVES A scoped pilot proposal with phases, risks, success criteria, and open decisions ## Maya fills in the form The recipe carries the method 01 · SUPPLY THE BUSINESS CONTEXT ### Maya fills in the form The recipe carries the method Maya selects Create Project Proposal and supplies the goal, audience, constraints, and decision owners. The recipe defines what to investigate and what the proposal must contain. She can run the saved process without installing a CLI or reconstructing the author’s command and prompt. PROJECT-PROPOSAL@3 · SAMPLE INPUTS The same inputs appear in the form below ## Reuse the foundations Agree on the first useful scope 02 · INVESTIGATE AND RESOLVE THE CHOICE ### Reuse the foundations Agree on the first useful scope GenieWorks finds that existing account provisioning can support the goal. Public signup would also introduce billing and abuse-control work. That changes the proposal. It brings the pilot-versus-public decision to the team, then carries the agreed constraints into the result. APP-73 · Proposal history - MMaya Product leadRan Create Project Proposal with the onboarding goal and enterprise audience. - GenieWorks AI engineerWe can reuse provisioning. Public signup adds billing and abuse controls. Start with an invited enterprise pilot? - MMaya Product leadYes, pilot first. @Alex, can we keep our current account approval? - AAlex Engineering leadYes. Reuse provisioning and manual approval. Add guided setup after the invitation. - GenieWorks AI engineerProposal ready for review: invited-pilot scope, delivery phases, risks, and success criteria. Pilot customers, targets, and timing remain open. ## Inspect the proposal form and result 03 · REVIEW THE DELIVERABLE ### A proposal the team can make decisions from The form and output describe the same work. The result preserves the invited-pilot decision and the existing approval flow, with remaining choices still visible. The proposal is ready for team review. It does not approve implementation, set a launch date, or remove the existing account-approval step. ## Make a useful method available to the next teammate 04 · REUSE THE TEAM’S PROCESS ### The author defines the method The team can run it A versioned recipe keeps the input form, investigation, and expected output together. The next teammate can use the same method and inspect the run’s inputs, decisions, and result. ### Define the investigation Choose the inputs, project context, tools, decision points, and deliverable. ### Run the saved recipe Supply business inputs and answer consequential questions in the shared workflow. ### Inspect how it got there Keep the recipe version, inputs, decisions, and result with the run. ## Turn the agreed proposal into work CONTINUE THE WORK ### Turn the agreed proposal into work Use the accepted scope as the starting point for detailed specifications. Discuss your workflow Explore specification # Inside your tracker, connected to your systems — GenieWorks Page: [GenieWorks lives in Jira, GitLab, or GitHub](./integrations.html) ## Trackers & MCP TRACKERS & MCP ### GenieWorks lives in Jira, GitLab, or GitHub Your team’s decisions and live evidence stay together on the issue. GenieWorks investigates through the compatible connections your project authorizes. Follow the work ### Alex assigns the investigation inside the tracker Request, assignee, progress, and history stay with the issue WHAT THE TEAM RECEIVES A team-visible explanation of the delay, its sources, and the next step ## The repository cannot show today’s customer state 01 · ASK THE SYSTEM FOR FACTS ### The repository cannot show today’s customer state The source explains how export limits work. It cannot reveal Northstar’s current queue, active setting, or commercial entitlement. GenieWorks uses selected Sentry and Internal Operations MCP connections to bring those facts into the investigation before recommending a change. ### Sentry Are exports failing, or waiting? What do this morning’s traces show? ### Your internal service What limit is active for Northstar, and what does the customer’s plan allow? Compatible, authorized MCP connections selected for this project ## The next teammate can follow the reasoning 02 · KEEP EVIDENCE IN THE SHARED HISTORY ### The next teammate can follow the reasoning Alex’s assignment, the tool questions, and the returned evidence remain part of the same work. A teammate with access can see what was checked and why the recommendation changed. The investigation finds a configuration mismatch. It does not infer a worker defect simply because the ticket says exports are slow. Northstar · Issue comments & history - AAlex DeveloperAssigned to GenieWorks · Investigate slow exports for Northstar - GenieWorks AI engineerSentry: show this morning’s slow-export traces for Northstar. - Sentry MCP ObservabilitySix exports queued for 18–24 minutes. No worker errors in the retrieved traces. - GenieWorks AI engineerInternal Operations: compare Northstar’s active export limit with its entitlement. - Operations MCP Custom / internalActive limit: 1 export. Enterprise entitlement: 4. No intentional customer override recorded. - GenieWorks AI engineerThe active limit does not match the entitlement. Trace how it was set, then specify the correction and a regression check. The investigation has not changed production settings. ## Inspect the evidence behind the next step 03 · MAKE THE NEXT STEP EXPLAINABLE ### A mismatch to investigate A reason to investigate it The source of each fact stays explicit. Queue delay and a limit mismatch support the next investigation; the retrieved traces alone do not prove there are no worker defects anywhere. ### 6 exports waiting 18–24 minute queue delays; no worker errors in the retrieved traces. ### 1 active / 4 entitled The live customer setting differs from its plan. No intentional override is recorded. ### Trace the setting’s origin Establish the cause, agree the correction, and verify behavior against the entitlement. ## Keep the tracker and connections your team uses 04 · MAKE IT PART OF THE TEAM’S WORK ### Your tracker is already the shared workspace Keep conversations, agreed decisions, and handoffs where teammates can find them. Connect compatible observability and internal services for the evidence the repository cannot supply. JiraView product GitLabContact us GitHub IssuesContact us Use the project’s authorized MCP connections, including your custom/internal services. ## Carry the evidence into an agreement CONTINUE THE WORK ### Carry the evidence into an agreement See how an investigation becomes a bounded specification the team can review and implement. Explore specification # Your models and infrastructure — GenieWorks Page: [Your models and infrastructureOne engineering process](./deployment.html) ## Models & deployment MODELS & DEPLOYMENT ### Your models and infrastructure One engineering process Use your preferred compatible models, choose where execution runs, and keep the work inside your team’s tracker. GenieWorks carries the engineering workflow around those choices. Follow the work ### “Can we use our models and run in our environment?” WHAT THE TEAM RECEIVES A clear split between chosen execution and managed team coordination ## Choose models, approach, and execution location 01 · SEPARATE THE CHOICES ### Choose the setup around the work Model capability, the engineering method, and where execution runs are separate decisions. Your team can configure them together without moving the workflow into someone’s personal session. ### Use compatible connections Connect a supported provider endpoint or a compatible model you operate. Keep the choice in your team’s configuration. ### Define permitted approaches Set the model, reasoning depth, instructions, and tools for different kinds of work. ### Choose the environment Use hosted execution or customer-hosted execution that connects to the managed coordination service. Inspect the specification profiles ## Follow a task across the deployment boundary 02 · FOLLOW ONE TASK ### The environment changes The shared process carries through The tracker supplies the task. Managed coordination supplies the workflow and selected context. Execution uses the chosen model, then returns results and usage to the shared record. ### Your tracker Issue, team decisions, and agreed outcome ### GenieWorks Managed coordination, task queue, project context, and encrypted credentials ### Your environment Workflow execution, compatible models, and your compute Results and usage return to the managed service and shared team history ## Know what runs where 03 · MAKE OWNERSHIP EXPLICIT ### Customer-hosted execution Managed coordination Choosing your own execution environment does not mean the entire product is self-hosted. Coordination and encrypted credential storage stay managed; results and run records return to GenieWorks. **Deployment responsibilities** | Responsibility | Hosted execution | Customer-hosted execution | | --- | --- | --- | | Workflow execution | GenieWorks environment | Your environment | | Model connection | Selected compatible provider | Selected compatible provider or model you operate | | Task coordination and queue | Managed by GenieWorks | Managed by GenieWorks | | Encrypted credential storage | Managed by GenieWorks | Managed by GenieWorks | | Results, usage, and run records | Returned to the managed service | Returned to the managed service | ## Keep work and history available to the team 04 · RETURN THE RESULT TO THE TEAM ### The work stays shared when a laptop closes Teammates can see what is running, what needs input, and what finished. These product views demonstrate shared work and history; they do not prove the hosting location of a run. Explore shared execution ## Built for products people depend on A SHARED TEAM CAPABILITY ### Built for products people depend on This process fits teams evolving an existing product, where context, compatibility, review, and a clear definition of done matter. Early exploration that changes direction every hour is a weaker fit; team size is not the deciding factor. Feature tags: Your model choices · Your execution environment · Shared team context · Inspectable work. ## Bring your setup into the conversation CONTINUE THE WORK ### Bring your setup into the conversation Discuss your models, connectivity, and execution requirements. Work out the deployment boundary your team needs. Trackers and MCP connections Discuss your deployment # Proptech SaaS Case Study — GenieWorks Page: [Proptech SaaS Case Study](./case-study.html) ## Proptech SaaS Case Study Home ### Proptech SaaS Case Study A strong product team already using AI individually brought it into a shared delivery process. ## Proptech SaaS Case Study ### Why an Engineer’s mindset matters A strong, fast-moving Proptech SaaS team already used AI in daily development. GenieWorks brought that capability into a shared engineering process: investigated requests, clear agreements, and implementation reviewed against the intended outcome. Project context and decisions carried through the work, giving the whole team a common way to deliver with AI. ### Proptech SaaS Case Study About 4× estimated effective delivery capacity Same team size · Already using AI · March vs January Estimated delivery capacity Jan–Jun 2026 1.0×Jan 1.9×Feb 3.9×Mar 7.6×Apr 8.0×May 7.3×Jun January = 1× · Select a month to explore ## How the estimate was made ### What the 4× estimate means March compared with January 2026: 3.9× estimated effective delivery capacity, rounded to 4×. Team size was unchanged, and developers already used AI individually. The estimate combines recorded human work with estimated GenieWorks effort from delivered changes and specification work. This comparison covers the team’s customized adoption of GenieWorks. Missing worklogs and estimated effort limit it; it does not establish delivery speed, cost savings, or GenieWorks’ isolated contribution. Source: corrected CPO Delivery Impact Report, January–June 2026. Our ambition · We believe 5–10× is possible when teams gain both AI execution and stronger delivery discipline. This is an expectation, not a measured result. # Let’s Talk — GenieWorks Page: [Bring your team’swork into the conversation](./contact.html) ## Let’s Talk - GenieWorks - Let’s Talk LET’S TALK ### Bring your team’s work into the conversation Tell us what you want to improve. We’ll explore how GenieWorks fits your delivery process, models, and infrastructure. - Start with a real taskShow us where investigation, implementation, or review slows the team down. - Keep your team’s workflowJira, GitLab, or GitHub — with shared context and visible decisions. - Discuss your requirementsYour model choices, execution environment, and Enterprise needs. Prefer email? anton@genieworks.co ### Tell us about your team We’ll use these details to respond to your enquiry. Send message ### Thanks for reaching out Your message is on its way to our team. We’ll reply to the email address you provided. Back to GenieWorks