The husband's attorney retained a forensic accountant three weeks before the equitable distribution hearing. Your client, the lower-earning spouse raising two children, learned about it in a status conference. She looked at you. You already knew the question.
Can we get one too?
The forensic accountant's rate is $350 per hour. A contested high-asset divorce requires 30 to 60 hours of forensic work. That is $10,500 to $21,000 on top of legal fees your client is already struggling to cover. The marital estate has $900,000 in disclosed assets. The word "disclosed" is doing a lot of work in that sentence.
You tell her you will handle the financial analysis yourself. You mean it. But you also know what that means: nights and weekends with bank statements, spreadsheets, and a calculator, looking for the transfers and accounts that a trained forensic accountant would find in half the time.
This is the asymmetry that decides contested divorces. The side with better financial analysis wins. Not better lawyering. Better analysis. And small family law firms have been on the wrong side of that asymmetry for decades.
Private AI changes the math.
What you are missing in financial discovery
You know where to look. Every family law attorney who has handled a contested divorce knows the patterns. The problem is not knowledge. It is capacity.
The transfer that does not match. Husband sends $40,000 to his brother eight months before filing. The brother bought a boat. The husband says it was a loan repayment. There is no promissory note. You would catch this if you had time to cross-reference 24 months of bank statements line by line against the sworn financial affidavit. You do not have that time when you are carrying 35 active cases.
The account that was never disclosed. Wife's mandatory disclosures list three bank accounts and a 401(k). Her tax return shows interest income that does not match the disclosed accounts. There is a fourth account. Maybe a fifth. You would find it if you could run the numbers on every financial document in the file against every disclosure. That analysis takes a full day you do not have.
The spending pattern that changed. Six months before filing, the monthly credit card spend drops by $3,000. Where did that money go? Cash withdrawals increased. A new P.O. Box appeared on a bank statement. Dissipation of marital assets is a claim you can make, but only if you can document the pattern with specificity. Judges want numbers, not narrative.
The business valuation gap. Husband owns a 40 percent interest in an LLC. The operating agreement has a buyout formula. But the company's QuickBooks data tells a different story than the tax returns. Revenue is growing. The K-1 distributions do not reflect it. A formal business valuation costs $15,000 to $25,000. Your client cannot pay it. So you accept the number opposing counsel offers and hope it is close.
These are not obscure forensic puzzles. They are standard contested divorce issues that small firms handle every week. The difference between a good outcome and a bad one is whether you have the resources to catch what the other side is hiding.
What private AI makes possible
Dhakma Core is a self-contained appliance that sits in your office. Pre-configured. No IT department required. Dhakma installs it, supports it, and keeps it running. Your team uses a secure internal web interface. Here is what it changes for a family law practice.
Financial anomaly detection
Before: Your paralegal receives 24 months of bank statements, credit card statements, and brokerage account records. Half are scanned PDFs from opposing counsel's production. Some are barely legible. They review them manually, highlighting anything that looks unusual. They build a spreadsheet of large transactions. They miss the pattern because patterns require comparing hundreds of data points across multiple accounts simultaneously. A human reviewing line items sequentially will miss what a system reviewing all data points concurrently will catch.
With Dhakma Core: Feed in the complete financial record set. The system handles scanned PDFs, inconsistent formats, and the reality of what opposing counsel actually produces in discovery. It maps every transaction across every account chronologically. It flags transfers between accounts that do not appear on both sides. It identifies new payees that appeared in the six months before filing. It calculates monthly spending by category and highlights deviations from the historical baseline. It cross-references disclosed assets against the transaction record and flags accounts that received deposits but were not listed in mandatory disclosures.
What this unlocks: You walk into the equitable distribution hearing with a transaction-level analysis that opposing counsel's forensic accountant would charge $15,000 to produce. Your client paid nothing extra for it. It is part of your representation.
Asset tracing
Before: Your client says the husband has a rental property she found out about from a neighbor. You subpoena county records. You find the deed. It is held in an LLC. The LLC has a registered agent but no obvious connection to the husband. You would need to trace the formation documents, bank accounts, and tax filings to establish the connection. That paper trail takes 10 to 15 hours of paralegal time per entity.
With Dhakma Core: The system ingests every document in the case file. Tax returns, bank statements, business records, property records, correspondence. It cross-references names, addresses, EIN numbers, and account numbers across the entire document set. When an entity appears in one document, the system finds every other document where that entity, its principals, or its accounts appear. The connections surface automatically.
What this unlocks: You find the undisclosed brokerage account because a quarterly dividend payment appeared on a bank statement the husband thought no one would read past page 40. The system read every page.
Custody memo drafting
Before: Your attorney spends six hours drafting a parenting plan proposal and best-interest analysis for a contested custody matter. They pull from the GAL's report, school records, the psychologist's evaluation, and their own case notes. They organize the statutory best-interest factors and map the evidence to each factor. It is substantive legal work that cannot be delegated to a paralegal.
With Dhakma Core: The system ingests the guardian ad litem's report, psychological evaluations, school records, and your case notes. It builds your initial draft by organizing all key facts and quotes from the record under the correct statutory best-interest factors, creating a comprehensive annotated outline in minutes. The attorney provides the legal argument, strategic framing, and credibility assessments that only a human practitioner can deliver. Total time from case file to filed memo: cut in half.
What this unlocks: Your attorneys spend their time on courtroom strategy and client preparation instead of combing through 200-page GAL reports to find the three paragraphs that matter. The organized outline means nothing gets missed.
Lifestyle analysis for alimony
Before: Establishing the marital standard of living for a maintenance argument requires reconstructing the household budget from financial records. Credit card statements, bank withdrawals, mortgage payments, tuition bills, travel expenses. Your paralegal builds a spreadsheet by hand. It takes a full day per year of records analyzed. For a 15-year marriage, that analysis is often abbreviated because the time cost is prohibitive.
With Dhakma Core: The system categorizes every transaction across the marriage's financial records by expense type. Housing. Transportation. Education. Travel. Dining. Medical. Discretionary. It calculates monthly and annual totals by category. It produces the spending analysis that establishes the marital standard of living with the specificity that judges require for imputing income or setting long-term maintenance.
What this unlocks: Your alimony argument is built on transaction-level data, not estimates. When opposing counsel argues the marital standard of living was lower than you claim, you have the receipts. Every one of them.
The data is too sensitive for cloud AI
This section is brief because the analysis is obvious.
A family law case file contains Social Security numbers for both spouses and their minor children. Financial account numbers. Mental health treatment records. Substance abuse evaluations. Domestic violence documentation. Guardian ad litem reports discussing children by name. Custody evaluations with psychological testing results.
This is the most sensitive category of legal file in private practice. There is no close second.
ABA Model Rule 1.6(c) requires reasonable efforts to prevent unauthorized disclosure. Uploading a client's financial affidavit and custody evaluation to a cloud AI vendor does not meet that standard. It does not matter what the privacy policy says. The moment the data leaves your building, you have lost control of it.
Dhakma Core has no cloud connection. The air-gap switch is a physical hardware disconnect. No software vulnerability bridges a physical gap. Your client's SSNs, their children's information, and their financial records stay on the machine in your office.
One contested divorce pays for it
Your firm handles a contested divorce with $1.2 million in marital assets. Opposing counsel retained a forensic accountant. You used Dhakma Core. Your financial analysis identified $180,000 in undisclosed assets the husband transferred to family members over 18 months. The court included those transfers in the equitable distribution calculation. Your client's share increased by $90,000.
Your legal fee on the improved outcome more than covers the system's monthly cost. And you did not bill your client $15,000 for a forensic accountant. You offered forensic-level analysis as part of your representation. That is a competitive advantage no other small firm in your market has.
Now consider the referral effect. The client tells her friend going through a divorce. The friend calls your office. She says: "I heard you found the hidden accounts in Sarah's case." That reputation compounds.
See it on a closed case
Bring the financial records from a completed divorce. A case where equitable distribution is done and the numbers are final. Redact what you need to redact.
We run the financial documents through Dhakma Core on-site. You watch. Your paralegal compares the anomaly detection output against the financial analysis they performed by hand. The transfers they caught. The ones they missed.
If the system does not surface something your team did not find, we take the hardware home.
The side with better financial analysis wins the contested divorce. That used to mean the side with the bigger budget. Not anymore.