The numbers do not agree
Finance, marketing, and operations report different versions of performance, and leadership cannot tell which definition or period should drive the decision.
AI & operations consulting
Diagnostic / evidence / action
Find the gaps between your reports, your systems, and how work gets done.
We connect conversations with the underlying workflows, code, and data—then turn supported findings into a blocker map and a practical 30–60–90 plan.
Four weeks$10,000–$20,000 USDSenior-led
When to bring us in
You have a decision to make, but the pieces do not line up.
We agree on a focused question and follow the evidence across the relevant parts of the business—without treating the first explanation as the final answer.
Work stalls between teams
A project keeps slipping because the blocker sits across ownership, workflow, data, and system boundaries rather than inside one team.
An AI investment needs a reality check
The proposal is compelling, but nobody has verified whether the available data, process, permissions, and operating model can support it reliably.
The explanation does not match the system
A metric, policy, or operational belief sounds reasonable until it is checked against the code, configuration, source records, or actual sequence of work.
The engagement
A four-week investigation with a decision at the end.
We review the records your business already produces, interview the people responsible, and test the leading explanations against the relevant code, configurations, calculations, and read-only data.
A typical scope includes 10–20 stakeholder interviews and 10–20 detailed written discovery questions. The exact mix is agreed around the decision, source volume, and team structure.
Findings with evidence
Each material issue includes its supporting sources, confidence status, decision impact, limitations, and the next validation step.
Dependencies and blockers
A practical map of the workflows, systems, data, and responsibilities involved—showing where a problem originates and what must happen first.
A prioritized 30–60–90 plan
Recommended changes sequenced by impact, dependency, effort, owner role, and acceptance criteria rather than by opinion or departmental urgency.
An implementation brief
Requirements for the first recommended change: what it should do, what it depends on, and how the team will know it works.
An organized research record
An exportable evidence register and handover so the team can inspect the reasoning, revisit open questions, and continue the work.
Outcome
Your team knows what is supported, what remains uncertain, and what to do next.
How the month works
Evidence develops in the open—not as a surprise report in week four.
Week 1
Establish the question and sources
Define the decision, map the selected workflow, confirm access, and begin stakeholder interviews and written discovery.
Week 2
Investigate and test
Trace the leading explanations across conversations, code, configurations, calculations, and read-only system data. Review the first findings with the client.
Week 3
Define the changes
Turn supported findings into sequenced changes, ownership decisions, measurement requirements, and a specification for the first improvement.
Week 4
Review and hand over
Walk leadership and the delivery team through the evidence, unresolved questions, blocker map, and 30–60–90 day plan.
The four-week schedule starts once agreed access is available. A first readout is held at the end of week two so missing context and competing explanations can be addressed early.
Client work
Connecting reporting, customer experience, and internal systems.
In work with a publishing and e-commerce business, Blatt followed operational questions across conversations, code, reporting, and product records.
Anonymized engagement
Question
Why do operational and financial views lead to different decisions?
Sources
Leadership interviews, team conversations, reporting logic, product records, and system definitions.
Established
The investigation documented inconsistent cost treatment, mismatched reporting periods, and missing checks between product promises and product content.
Boundary
Findings were separated from hypotheses, and financial impact remained subject to corrected calculations and policy decisions.
Handover
The research was organized into an evidence register, prioritized actions, and engineering handoffs—not a claim of automatically realized savings.
Sources we may review
The evidence already exists. It is usually scattered across formats and owners.
The source set is selected for the agreed question. Access is read-only where appropriate and limited to the records, period, systems, and participants defined in the engagement scope.
01
Slack and internal messaging
02
Meeting transcripts
03
Documents and operating records
04
Detailed written discovery responses
05
Selected codebases and configurations
06
Read-only reporting and operational data
MemoryOS
Company context that can be traced back to its sources.
MemoryOS is Blatt's proprietary research infrastructure. It helps organize conversations and documents around people, topics, decisions, and dependencies while retaining references to the original records.
Role in the engagement
It supports cross-source investigation and preserves the research context. Findings are still reviewed against source records, technical mechanisms, limitations, and professional judgment.
Client-controlled GCP
Processing takes place remotely in a Google Cloud environment provisioned and controlled by your company.
Client-provided model access
Language-model requests use API credentials supplied by your company and an agreed provider configuration.
No-retention settings
The selected provider is configured not to retain request data. Approved requests may still be sent from GCP to that provider's API.
Source access, authorized users, retention, credential removal, and model-provider settings are documented for the engagement before company records are reviewed.
Senior-led delivery
Technical depth without the junior handoff.
Led by Blatt's CTO
The engagement stays senior-led from kickoff through the final review, combining AI systems architecture with an MBA-level business perspective.
One connected investigation
Operational interviews, data analysis, and technical verification are kept in one line of reasoning instead of being split across disconnected workstreams.
Investment
One focused engagement. A defined scope and price.
$10,000–$20,000 USD for four weeks. The fixed price reflects source volume, technical depth, workflow complexity, and whether one focused implementation is included.
- Diagnostic is a standalone project
- Infrastructure and model costs identified in advance
- Further implementation and MemoryOS access scoped separately
Frequently asked questions
Before the work begins.
Start with the decision
What is your team struggling to understand or change?
Tell us where the numbers conflict, the work stalls, or the next AI investment feels uncertain. We'll discuss whether a focused four weeks engagement fits.