- An estimated 2.44 million people were injured in U.S. crashes in 2023, and nearly half of claimants who hire a lawyer do so within days of the accident.
- Intake usually pays back first: in Clio's 2024 study of 500 firms, only 40% answered the phone and 48% could not be reached at all.
- During treatment and records collection, AI keeps clients on schedule and tracks every outstanding record and bill, so gaps do not weaken the claim.
- AI can prepare the medical chronology and a first draft of the demand, provided every entry cites its source page and a lawyer reviews the result.
- AI should not give legal advice, decide whether to accept a case or negotiate with insurers.
An estimated 2.44 million people were injured in motor vehicle crashes in the United States in 2023 (NHTSA), and many look for a lawyer quickly. In a 2026 survey by the Insurance Research Council (an insurer-funded research body), nearly half of auto injury claimants who hired an attorney did so within days of the accident, and 80% within weeks.
Many firms are not set up to respond at that speed. When researchers for Clio's 2024 Legal Trends Report contacted 500 U.S. law firms as prospective clients, only 40% answered the phone, and almost half (48%) neither answered nor returned the call. In personal injury, an unanswered enquiry is usually a case signed elsewhere. This guide follows a case through each stage where time is lost, sets out what AI can take on and what must remain with a lawyer, and shows what it costs.
The six stages of a case
- IntakeThe firm records the facts and the client signs the fee agreement and authorizations.
- TreatmentThe client receives medical care, often over several months.
- PackagingThe firm obtains every outstanding record and bill.
- DemandThe firm sends the insurer a demand supported by the records.
- NegotiationThe firm and the adjuster negotiate a settlement figure.
- Settlement or litigationThe case settles, or the firm files suit.
AI is useful in the first four stages, where the work is repetitive and time-sensitive. The last two depend on a lawyer's judgment.
Intake: respond to every enquiry
Accidents do not happen during office hours, and many enquiries now arrive by text or WhatsApp rather than by phone. An agent can conduct the same conversation on each channel, and screen out enquiries the firm would not accept: callers who were at fault, were not injured or have already retained a lawyer.
A few details decide whether a case is viable, and a generic chatbot will miss them. The agent should record the accident date against the state's limitation period. It should establish which insurer pays first: in no-fault states such as Florida, Michigan and New York, the client's own personal injury protection (PIP) coverage pays initial medical bills, and in every state firms usually notify the client's own insurer to preserve a possible uninsured or underinsured motorist (UM/UIM) claim. It should also ask how treatment will be paid for, since a client without health insurance may need care under a letter of protection.
The attorney consultation remains a human step. The agent records the facts and books the consultation at 11 p.m.; the attorney reviews the fee agreement with the client the following morning.
Intake checklist
Whether intake is handled by staff or by an agent, each new matter should leave the first conversation with the following recorded.
| Area | Record |
|---|---|
| The accident | ☐ Date, time and location ☐ Police report number ☐ Party at fault and any citations ☐ Photographs of vehicles, scene and injuries ☐ Witness names and contact details |
| Injuries and treatment | ☐ Emergency or urgent care, and the date ☐ Current symptoms ☐ Providers seen to date ☐ Prior injuries to the same part of the body |
| Insurance and payment | ☐ At-fault driver's insurer and claim number ☐ Client's own policy, including PIP or MedPay ☐ Uninsured or underinsured driver involved ☐ Health insurance, or need for a letter of protection |
| Deadlines and documents | ☐ Limitation date under state law ☐ Contingency fee agreement signed ☐ HIPAA-compliant medical authorization signed ☐ Attorney consultation scheduled ☐ Written guidance on appointments, exercises and social media |
| Client concerns | ☐ Lost income ☐ Vehicle repair or rental ☐ Outstanding bills or collection calls |
Treatment: prevent gaps in care
Treatment records form the core evidence of the demand, and insurers examine them for gaps. If a client stops physical therapy for six weeks, the insurer is likely to argue that the injury was minor or had another cause. Monitoring every client is difficult for a case manager with 80 or more files. An AI agent can:
- Contact clients before appointments and follow up afterwards.
- Flag any client who has gone more than two weeks without a visit.
- Answer routine status questions from the case management system.
Packaging: obtain every record
Obtaining records is often slow because providers handle requests in very different ways. A study of 83 top-ranked U.S. hospitals published in JAMA Network Open found that only 53% of their request forms offered the option of the complete record, and that quoted fees for the same 200-page record ranged from nothing to $541.50.
Requests require a HIPAA-compliant authorization from the client (45 CFR 164.508), and records often need supporting affidavits to be used at trial; in Texas, for example, under Section 18.001 of the Civil Practice and Remedies Code. An agent can track every record, bill and affidavit by provider, send follow-ups on a schedule, and compare what has arrived against the treatment log, so a missing month of therapy is caught before the demand rather than by the insurer.
Demand package: prepare the first draft
Most of the time spent on a demand goes into reading records, not writing. A standard medical chronology has one row for each date of service:
| Column | Content |
|---|---|
| Date of service | In one consistent format, so the table can be sorted |
| Provider | Named consistently throughout |
| Summary | Complaints, diagnosis, treatment, work restrictions, recommendations and any statement on causation |
| Prior history | Any pre-existing condition noted in the record |
| Page reference | The page or Bates number where the entry appears |
| Notes | Relevance to the claim |
Two principles apply when AI prepares it. Every entry must cite its source page, so it can be checked. Key passages should be quoted rather than paraphrased, leaving interpretation to the lawyer. An AI first pass can also flag what insurers look for, such as symptoms that change between visits, gaps in treatment and consistently maximal pain scores, so the lawyer can address them before the demand is sent.
Where AI should stay out
AI should not give legal advice, decide whether to accept a case, negotiate with insurers or send anything without review. Purpose-built tools are not exempt from error: a Stanford study published in the Journal of Empirical Legal Studies in 2025 found that the AI research tools from LexisNexis and Thomson Reuters each produced incorrect or misleading answers more than 17% of the time.
ABA Formal Opinion 512 (2024) requires lawyers using generative AI to protect confidentiality and supervise its output, and state bar advertising rules apply to anything the agent says. Outbound calls using AI-generated voices also require the recipient's prior consent under the Telephone Consumer Protection Act (FCC 24-17, 2024).
What AI costs a personal injury firm
| Option | Typical cost | What it covers |
|---|---|---|
| In-house intake employee | About $3,600 to $5,600 a month in wages, before benefits | Business hours, unless you staff evenings and weekends |
| Human answering service | $250 to $2,100 a month | Calls answered by people, billed per minute or per call |
| AI intake receptionist | About $150 to $800 a month for 75 to 500 calls | Answers at any hour, screens enquiries and books consultations |
| AI medical chronology tool | From $25 plus $0.01 a page | A first-draft chronology that a lawyer still checks |
| Records and demand platform | Quoted per case | Records review and demand drafting |
| Custom-built agent | From $3,000 to build, plus usage | Intake, follow-ups and records tracking built around your process |
The wage range is the May 2025 Bureau of Labor Statistics mean pay for receptionists ($42,950 a year) and paralegals ($67,500 a year) in legal services. Answering-service prices come from Ruby's and Smith.ai's published plans, AI intake from Smith.ai's AI receptionist plans, and the chronology price from CaseMark. EvenUp charges one price per case, and Supio quotes per case or for unlimited firm access, so ask both for a quote based on your caseload. Usage on a custom agent covers the AI model and phone minutes; voice platforms such as Retell AI list $0.07 to $0.31 a minute.
The quickest test is to compare the monthly cost with your average fee from one signed case. If the fee is larger, a single case a month that would otherwise have gone to voicemail covers the cost.
Buy or build
| Buy a ready-made tool | Build your own agent | |
|---|---|---|
| Examples | Smith.ai for intake; EvenUp, Supio or Eve for records and demands | An agent built around your intake script and case management system |
| Cost | About $150 to $800 a month for AI intake; demand platforms quote per case | From $3,000 to build, plus usage |
| Time to start | Faster | Most custom agents go live in 4 to 8 weeks |
| Fit with your process | Your team adapts to the tool | The tool follows your process |
| Client data | Held by the vendor, under its terms | In your own accounts, with AI providers you approve |
Whichever route you choose, get clear answers to four questions:
- Is client data kept or used to train AI, and will the provider sign a confidentiality or data processing agreement?
- Can you export all of your data at any time?
- Does it write directly into your case management system, such as Filevine, SmartAdvocate, Litify, CASEpeer or Clio?
- Can you start on a monthly contract before committing long term?
How Softpers builds a custom agent
Softpers builds AI agents around the way your firm already works, so your team does not have to change its process to fit a tool. For a personal injury firm, that usually starts with intake and extends to the stages that follow:
- Intake: an agent answers calls, texts and web chats at any hour, asks your best intake specialist's questions, screens out cases you would not accept and books the attorney consultation.
- Your case management system: each new matter, its documents and its deadlines are written straight into the system you already use, so nobody re-keys them.
- Treatment and records: the agent checks in with clients between appointments, flags gaps in care and chases outstanding records and bills by provider.
- Lawyers in control: every action is logged, approval rules decide what the agent may send on its own, and nothing reaches an insurer without a lawyer's review.
AI agent projects start from $3,000. Each one begins with a free 30-minute discovery call and a fixed-price pilot of 1 to 2 weeks, so you see a working first version before committing to the full build. The build takes 2 to 6 weeks with weekly demos, launch takes about a week, and a 30-day warranty follows; most agents go live in 4 to 8 weeks. Client data stays in your own accounts, the AI models are never trained on it, NDAs and data processing agreements are standard, and the code, prompts and documentation are 100% yours.
How to start
Smaller firms have the most to gain: in the American Bar Association's 2024 Legal Technology Survey, only 18% of solo practitioners used AI, compared with 46% of lawyers at firms of 100 or more.
- Measure what is missedCount a month of after-hours, abandoned and unanswered enquiries.
- Settle the workflow and the dataAgree on one case management system, and confirm your contract lets you export all records.
- Start with intakeBuild it on your best specialist's questions, then extend to treatment and records tracking.
- Adopt an AI policySet out what AI may do, what data it may access and who reviews its work.
Can an AI intake agent sign up clients?
It can collect the intake details, send the contingency fee agreement and medical authorizations for e-signature, and book the attorney consultation. The decision to accept the case, and any advice about the claim, remain with a lawyer.
Is it ethical for a law firm to use AI with clients?
Yes, with supervision. ABA Formal Opinion 512 (2024) requires lawyers who use generative AI to protect client confidentiality, understand the tool's limitations and review its output. State bar rules may add further requirements.
Will callers know they are speaking with AI?
They should. In Clio's 2024 research, 51% of prospective clients said a chatbot can be a helpful starting point, but three in five would use one only if they could switch to a person. The agent should identify itself at the start and offer a person at any point.
Can AI write a personal injury demand letter?
It can prepare the first draft. AI can build the medical chronology, total the medical bills and draft the demand narrative from the records, citing a source page for every fact. A lawyer should check each fact against the records, decide the amount and approve the letter before it goes to the insurer.
How much does AI cost for a personal injury law firm?
An AI intake receptionist costs about $150 to $800 a month for 75 to 500 calls, compared with $3,600 to $5,600 a month in wages for an intake employee. Records and demand platforms such as EvenUp and Supio quote per case. A custom agent built by Softpers starts from $3,000, plus usage such as AI and phone minutes.
- Traffic Safety Facts 2023 Data: Summary of Motor Vehicle Traffic Crashes, DOT HS 813 762 (National Highway Traffic Safety Administration, 2025)
- Public Opinions on Attorney Involvement in Auto Insurance Claims: 2026 (Insurance Research Council, reported by Insurance Business, 2026)
- 2024 Clio Legal Trends Report: Fixing the first impression problem for law firms (Illinois Supreme Court Commission on Professionalism, 2024)
- ABA tech report finds that AI adoption is growing, but some are hesitant (ABA Journal, on the 2024 ABA Legal Technology Survey Report, 2025)
- Assessment of US Hospital Compliance With Regulations for Patients' Requests for Medical Records (Lye et al., JAMA Network Open, 2018)
- Hallucination-Free? Assessing the Reliability of Leading AI Legal Research Tools (Magesh et al., Journal of Empirical Legal Studies, 2025)
- Formal Opinion 512: Generative Artificial Intelligence Tools (American Bar Association, 2024)
- Declaratory Ruling on AI-generated voices under the TCPA, FCC 24-17 (Federal Communications Commission, 2024)
- 45 CFR 164.508: Uses and disclosures for which an authorization is required (eCFR)
- Texas Civil Practice and Remedies Code, Chapter 18 (Texas Legislature)
- Occupational Employment and Wage Statistics: Legal Services, May 2025 (U.S. Bureau of Labor Statistics, 2026)
- AI Receptionist pricing (Smith.ai, 2026)
- Virtual Receptionists pricing (Smith.ai, 2026)
- Virtual receptionist pricing (Ruby, 2026)
- Medical record chronology (CaseMark, 2026)
- EvenUp launches case-based pricing (EvenUp press release, 2025)
- Supio pricing plans (PricingSaaS, 2026)
- Retell AI pricing (Retell AI, 2026)
