Some work should be done by hand simply be ready.
AI workflows for recurring business decisions.
Built inside your AI tools. Checked, tracked, and improved every run.
If AI can do the work, why are you still working on it?
You have systems, dashboards, reports, and copilots.
AI can now produce serious business deliverables.
Someone still has to do the work: read what changed, judge what matters, and prepare what needs action.
What if Monday’s work was already waiting for review?
Not a blank page. Not a prompt to start from.
The facts changed overnight: cash moved, the forecast shifted, a bank file arrived.
By the time the day starts, the analysis is refreshed, checked, and ready for the decision.
Plausible is not enough.
A review can look finished and still be wrong.
Better models help. They still do not arrive knowing what “right” means in your company.
So we build around your own work instead: the sources, checks, exceptions, and decision rules.
Start with a working folder.
Not a platform migration. Use the files, data, examples, and reviewer notes already used for a recurring decision.
That work becomes a repeatable skill: inputs, goals, checks, and outputs.
The workflow runs the recurring work for you.
--- name: treasury-weekly-review description: Turn the weekly treasury pack into a committee-ready memo with numeric traceability. --- # Treasury Weekly Review ## Operating Mode Delivery only. Hard gates fail loudly; no silent fixes. ## Inputs treasury_pack_week_19.xlsx rules-thresholds.json reviewer-notes.md ## Run Rules Reconcile cash and forecast totals. Stress liquidity using approved thresholds. Flag exceptions above policy limits. ## Output Committee memo: action, numbers, source notes. Traceability report: every number tied to source.
Treasury weekly review
Work is done for you: from data and dashboards to recommendations.
Then prove it’s right.
A good-looking review is not enough.
We run it again and again, with and without the workflow layer. It holds where AI alone drifts.
The proof is repeated runs, critical checks, measured consistency, and the time saved.
One workflow. Then another.
After the first job, you keep more than the output.
The workflow runs again, records cost and time saved, checks quality, and shows whether the standard is holding.
As more workflows are added, management gets one view of the work AI is actually doing.
For teams using AI but still doing the work.
Pick one recurring business decision you need to prepare for.
In two to three weeks, we prove the work behind it can be ready before you ask.
Fixed first scope. Continue only where value is proven. You keep what we build.
Start with one decisionOne recurring decision. Priced after we understand the decision, data, and delivery environment.
Eclipsai builds managed AI workflows for recurring business work, proven through repeated runs, checks, and run records. Led by Chip Alexandru, 20+ years in strategy consulting at BCG, PwC, and Accenture.
Questions before you bring one decision
Practical answers for teams considering a fixed first scope.
What do we get after the first proof?
You get the work behind one recurring business decision: the recommendation, supporting analysis, source evidence, and backup material. You also get the skill, run records, and output history needed to recreate and improve the work in future cycles.
What kind of work is this for?
Recurring internal business work where a team prepares facts, exceptions, analysis, and a recommendation before a decision. Good examples include commercial reviews, finance reviews, management notes, QBR packs, category reviews, account briefs, and market scans.
What is not a good fit?
Work that is not based on facts, sources, or analysis is not a good first fit. We also do not start with external reporting, audit, legal, HR, customer-facing decisions, regulatory filings, or workflows where an unchecked error would create unacceptable risk.
How does Eclipsai build the workflow?
We start with the decision: its inputs, outputs, quality bar, and expectations. The work behind it becomes a workflow standard: required files, source maps, gates, critical checks, evaluation criteria, and a run ledger.
Does the workflow make the decision?
No. The workflow prepares the work behind the decision. The decision owner remains responsible for reviewing the output, checking the evidence, and approving the final decision.
How do we know the output is right?
Before the proof starts, we agree what “good enough to support the decision” means. Typical checks include correct facts, source traceability, missing-data flags, exception handling, defensible recommendation, and reviewer confidence. The workflow can also be compared against direct AI use without the workflow layer.
What do you need from us?
One named recurring decision, the source files normally used, and ideally four to five prior examples of the report, review, pack, or memo. We also need the people closest to the decision quality for one scoping meeting and two joint review sessions.
Where does the workflow run?
The first workflow is built for Codex or Claude Desktop. The preferred setup is to run in your own approved environment. Broader Microsoft, AWS, or other platform implementation can be assessed later.
What does the first proof cost?
The first proof is fixed-scope and priced after we understand the decision, data, and delivery environment. It covers one recurring decision workflow. Multiple workflows, system integration, broad platform rollout, stakeholder training, and ongoing maintenance are separate.
What happens if the proof does not work?
If the workflow does not meet the agreed quality standard for the agreed one-decision scope, we keep improving within that scope or refund the proof fee. If refunded, the workflow is not used.
What happens after the first proof?
If the workflow proves value, you can maintain it internally, ask Eclipsai to maintain and improve it, or expand into more recurring decisions, each with its own skill, run records, checks, cost, and review status.
Start with one decision
Send a short note with the recurring decision or review you want to test.
Include the next cycle date if you already know it.