AI-enabled business intelligence
One ontology for everything your business knows about its data.
>What is our total funds under management?
Tested across organisations from elite sport to regulated finance.
- lower model cost
- 97%
- faster answers
- 3.4×
- questions correct
- 59 of 60
01 / Ontology
Your business rules, written once
What an active member is. Which accounts sit outside funds under management. Which table replaced which.
An ontology is the record of how your business defines its numbers. Akela keeps it in one place and reads it before every answer.
rules/house-accounts-are-not-members.md
- type:
- Rule
- tables:
- dim_cust, dim_acct, fact_acct_bal_mth, fact_txn
House accounts are not members
The business keeps some of its own seed money in the funds, in accounts held by house customers.
Leave house accounts out of funds under management, member counts, average balances and flows.
What is our total funds under management?
- With the ontology
- $5.19 billion
- Without it
- $5.38 billion
02 / Results
Faster, cheaper, and right
- Model cost$2,500$80
- Time spent waiting15 hours4.4 hours
- Model calls14,1006,400
- Tokens199 million94 million
Measured on a production warehouse at a regulated finance organisation.
Across 60 business questions, Akela answered 59 correctly. The same model without the ontology answered 38.
- With the ontology
- 59 / 60
- Without it
- 38 / 60
Cost, time, calls and tokens come from running the same 68 benchmark questions through the earlier multi-stage design and through the ontology design. Cost with the ontology is estimated from token use at published model prices. Accuracy was scored on a 60-question benchmark against our demonstration warehouse, by running each answer's query and comparing the rows with a known result.
03 / Dictionary
Every definition has one home
Akela centralises data definitions and governance in a single dictionary. Across domains. Across processes.
Every entry is checked against your live tables. Every change is versioned. The dictionary improves as your team asks, corrects and approves.
Live tables
- dim_acct
- dim_fund
- fact_acct_bal_mth
- dim_cust
- fact_txn
- fact_cmpgn_send
- fact_cmpgn_resp
- fact_svc_case
Checked against live tables
- Funds under managementMetric · Funds
- House accounts are not membersRule · Members
- Active memberTerm · Members
- Net flowsMetric · Flows
- Campaign response rateMetric · Marketing
- Service caseEntity · Service
Your team
- Asks
- Corrects
- Approves
04 / Query
Ask across the whole business
Funds, members, flows, marketing, service. Akela knows how they join and flags the join that would double-count.
No SQL required. Every insight delivered and visualised.
An example
Show funds under management by fund as a bar chart
Searching the knowledge
Show the loan book by region as a bar chart
- Plans the work6 s
- Works out and runs the queries24 s
- Draws the chart12 s
- Checks the chart5 s
- Writes the answer7 s
- Looks up the ontology4 s
- Queries the warehouse6 s
- Draws the chart and answers6 s
05 / How it works
The thinking is done before you ask
Other AI analysts plan, explore, query, check and rewrite on every question. Akela did that work once. It lives in the ontology.
So a smaller model answers in 6 calls where a larger one took 14. The same question, a third of the time, 3% of the cost.
Put Akela on your hardest data
We are letting people in a few at a time. Leave your email and we will send an invitation when there is room.