Quick answer. Across 58 postings, 53 were hourly and 5 were fixed price. Zero fixed-price projects exceeded $5,000. Twenty-six were contract-to-hire and 42 were for engagements over six months. The overwhelming majority were requests for a person to perform work by hand, at $3 to $25 an hour or $1,000 to $2,000 a month full time. Law firms in this market are not shopping for software. They are shopping for a body, and any software proposal is silently compared against a $12,000 to $24,000 annual salary.
There is a lot of writing about what law firms should automate. Most of it is written by people selling automation, using pain points collected from other people selling automation.
Job postings are different. A firm writing a job ad has no incentive to make its operations sound sophisticated. It has every incentive to describe the work precisely, because a bad description means bad applicants. The result is the most honest inventory of manual legal work available anywhere, published voluntarily, in public, for free.
So we read all of them.
Method, and its limits
One search for "Clio" on Upwork, late July 2026, sorted by relevance, both result pages, 58 postings total. A handful appeared twice. We read every posting in full, including the ones where Clio was one line in a long list of requirements, and classified each by what the firm was actually buying.
The honest caveats: one platform, one keyword, one snapshot. Upwork skews toward smaller firms and remote contract work, so this is not the market for a 300-attorney firm's procurement. Larger firms hire through different channels and buy through different processes. The counts by category are our classification and someone else might draw a line or two differently. The rate and volume figures below are taken directly from the postings and the platform's own filter counts, so those are exact.
The shape of the market, in numbers
Straight from the platform's filters:
- 53 hourly, 5 fixed price. Of the fixed-price jobs: two under $500, three between $1,000 and $5,000, and zero above $5,000.
- 26 of 58 contract-to-hire.
- 42 of 58 for engagements longer than six months. Eight were under a month.
- 38 wanted more than 30 hours a week. Five wanted fewer.
- 51 of 58 payment-verified. 34 of the clients had made 10 or more prior hires.
- 23 marked "expert" level, 35 "intermediate", zero entry level.
The single most informative number in that list is the first one. There is no fixed-price project above $5,000 anywhere in the set. Whatever these firms are buying, it is not a defined piece of work with a defined price.
Our classification of what they were buying:
- Around 33 postings: staffing. Paralegals, legal assistants, intake specialists, case managers, bookkeepers, virtual assistants, executive assistants, operations coordinators. Clio appears in the requirements, not in the job.
- 7 postings: Clio implementation consulting. Grow and Manage configuration, workflow and automation setup, document templates, third-party integrations, staff training. Advertised at $13 to $55 an hour.
- 4 postings: marketing. Meta ads, SEO, account management, where Clio Grow is one of the systems they will touch.
- 4 postings: engineering. Of which one paid $5 to $15 an hour, one was $1,000 fixed from an unverified client, one was a website build, and one was a genuine full-time integration engineer role.
Four engineering postings out of 58, and only one of them a serious technical scope. That one was an integration engineer owning the layer between a Salesforce environment and client CRMs including Litify, Filevine, SmartAdvocate, Clio, and CASEpeer, with OAuth flows, AWS Lambda, and per-client schema mapping. It was a headcount requisition with 50-plus applicants, not a project.
The inventory: what firms are paying humans to do
This is the part worth keeping. Consolidated from the postings, here is what the work actually is.
Intake and lead handling. Answer inbound calls, conduct the intake interview, qualify against firm criteria, schedule the consultation, follow up by phone, text, and email, keep the CRM current. One personal injury firm went further and published its desired automation map: stage service levels with named owners, attorney-review escalation, a medical malpractice fast-lane and turndown path, fee agreement e-signature reminders, document request cadences, required-field and source-capture fixes, and the Clio Grow to Clio Manage handoff. That is a specification, written by a buyer, given away for free.
Matter opening and file structure. Open and structure the matter, collect signed retainers and authorizations, apply the naming convention, save documents arriving by email into the right client folder with the right filename. Multiple postings specified "following internal filing and naming conventions" as a core duty. This is a human being used as a file router.
Records acquisition. Request and chase medical records and bills, order police reports, file freedom of information requests, coordinate property damage claims, open insurance claims, chase carrier claim files, send footage preservation letters. One New York plaintiff firm described this as its own dedicated function with hard service levels. We wrote that one up separately in medical records retrieval: vendor, hire, or automate.
Deadline calculation and calendaring. Calculate and calendar civil litigation deadlines, track court dates, monitor immigration court and agency deadlines, watch statutes, calendar a 90-day municipal notice of claim. Rules-based date arithmetic, performed by people, with case-ending consequences when it goes wrong.
Document assembly. Draft from firm templates, prepare filing packages, assemble exhibit lists with indexing and pagination and certificates of service, build medical chronologies and damages summaries, prepare demand packages. Immigration work is the extreme case: postings listed I-589, I-130, I-485, I-765, I-131, I-751, I-601A, N-400, EOIR-42A and 42B by form number, all assembled by hand.
Billing operations. Create draft invoices, circulate them for attorney review and signoff, chase the approvals, issue, record payments arriving by wire, check, and various processors, calculate revenue-share payouts, chase late clients. One ten-lawyer firm published its volumes: 60 to 70 invoices a month across more than 20 timekeepers.
Trust accounting. Monthly three-way reconciliation of client ledgers, trust ledger, and bank statement, reconciliation between the practice management system and the payment processor, and audit-ready record keeping. Covered separately in why three-way reconciliation still needs a human.
Client communication. Regular status updates on case progress, routine question handling, knowing when to escalate to an attorney. Several postings framed this as the main reason for the hire.
System integration by hand. Postings named the systems that must talk to each other: Clio with QuickBooks, LawPay, NetDocuments, Gavel, Dropbox, Microsoft 365, Outlook, e-signature, payment processors, Zapier, Salesforce, HubSpot, GoHighLevel, Lead Docket, RingCentral, Docketwise, and CallRail. Where the integration does not exist, a person is the integration.
The rate card
Published rates, taken directly from the postings:
- Legal admin, intake, and paralegal support: $3 to $25 an hour, clustering at $10 to $15.
- Full-time offshore intake specialist: $1,000 to $1,050 a month.
- Full-time offshore paralegal with case management system experience: $1,400 to $2,000 a month.
- Law firm billing and bookkeeping clerk: $4 to $20 an hour.
- Clio implementation specialist: $13 to $55 an hour.
- Expert developer integrating Clio, NetDocuments, Gavel, Zapier, and QuickBooks: $5 to $15 an hour, 30-plus hours a week.
Round the full-time figure: $12,000 to $24,000 a year for one person doing this work.
The finding
Law firms in this segment do not buy software for operational problems. They hire.
That is a rational preference and it deserves to be taken seriously rather than treated as a failure of education. Hiring has a known shape: a monthly cost, a monthly exit, and a person who can absorb an undefined scope from day one. Software requires the scope be defined before anything can be built, and defining the scope is itself the hard part when nobody in the firm has ever written down what the week contains.
It has a direct consequence for anyone selling to these firms, ourselves included. When we send a proposal, it is not compared against a Clio licence. It is compared against a salary. A five-figure build lands next to "I could hire someone for a year and a half for that," and that comparison is fair. Software's advantage is not that it is cheaper on day one. It is that its cost does not grow with volume and its knowledge does not resign.
Which of it a machine can honestly take
Splitting the inventory three ways, without flattering our own position.
Automates cleanly. Deterministic, rules-based, high volume, no judgment required.
- Deadline calculation from a trigger date and a rule set
- Matter creation and file structure from intake data
- Document naming and routing into matter folders
- Request logging, follow-up scheduling, and escalation
- Status updates generated from matter state
- Reconciliation comparison and exception flagging
- First-draft assembly from firm templates
Assists, with a human in the loop. A model produces a draft that a person checks. Real savings, and the review step is not optional.
- Medical chronologies and damages summaries from records already held
- Deposition summaries with citations back to page and line
- Correspondence and status letters
- Finding the one clause in the long contract
- Intake summarisation and initial routing
Stays human. Anyone who tells you otherwise has something to sell.
- The intake call with someone in the worst week of their life
- The precinct that has not posted the report and only answers on certain days
- Deciding whether an incomplete record is worth chasing
- Every judgment call about what a discrepancy or a gap actually means
- The professional responsibility that attaches to any of it
Roughly a third, a third, and a third. The middle third is where every honest conversation about legal AI should start and where most of them never get to, because the first third is easier to demo and the last third is easier to dismiss.
We're running this again, in the open. This inventory came from reading job postings from the outside. We're now asking firms directly: where does the workday actually go, and what's still stuck being manual. Anonymized, no pitch attached, full results back to every participant. Take the 2026 Legal Tech Research survey →
How to use this at your own firm
Write your own job posting for the role you would hire if money were no object. Do not post it. Just write it, listing every duty in the detail these firms used.
Then go through it line by line against the three categories above. The lines in the first category are your automation backlog, ranked by how many times a week they happen. The lines in the third are what you should actually be hiring for. The lines in the middle are the conversation worth having with someone technical.
Most firms discover two things doing this. The role they were about to hire is really three roles wearing one job title. And a surprising share of the first category is not automated because nobody ever wrote it down, rather than because it is hard.
Where Oktopeak fits
We build the first category and the plumbing under the second, on top of the systems firms already run. Clio stays your system of record; we build against its API, and we open-sourced our Clio and MyCase connectors so you can read the integration code before you talk to us.
We do not sell you a person, and for a meaningful share of that inventory a person is the right answer. If your problem is the third category, hire. If your problem is that someone with a law degree is spending Tuesday renaming PDFs, that is ours.
The first conversation is a free 30-minute review where you walk us through the week and we tell you which bucket each piece falls into, including the parts we cannot help with. Worst case you leave with the list.
