Make the Software You Own Earn Its Keep
How a Florida law firm can find, check, and use the AI capabilities already inside its current software before buying another license.
A former Big Law corporate equity partner at Steptoe (NYC Managing Partner) and Seyfarth Shaw • Florida-focused • Independent consultant • No software to sell
Find a useful starting point before the next sales call.
Get the firm AI-readiness checklist, tool-approval scorecard, five workflows worth testing, daily verification checklist, and the Small Law Firm AI Workflow Brief.
Get the Free Florida AI Starter Playbook →Last reviewed: September 30, 2026. This page is updated when common software, licensing, data protections, Florida guidance, or linked resources materially change.
A small law firm with fourteen software subscriptions and no inventory does not need a fifteenth subscription. It needs a flashlight.
AI is already appearing inside email, documents, meetings, legal research, practice management, intake, PDF, transcription, billing, and file-storage systems. Some features are included. Some cost extra. Some are enabled. Some are hidden behind an administrator setting. Some are being used through personal accounts while the partners believe the firm is “waiting on AI.”
The first job is not shopping. It is finding out what the firm owns, what those products can do, what they can reach, what the firm has approved, and whether any feature solves a recurring problem. Only then does a new purchase become a business decision instead of a reflex.
The short version
Six rules for making current software useful.
- Inventory the exact products, plans, licenses, administrators, renewal dates, and active users.
- Separate what is paid for, enabled, approved, and actually used. Those are four different lists.
- Match a feature to a defined workflow problem, not to a vendor demonstration.
- Check the exact account, settings, permissions, terms, data path, and reviewer before client information enters.
- Test one narrow use with fictional, public, or otherwise approved material.
- Measure the whole process, retire duplicates, and buy something new only when the evidence earns it.
On This Page
Your Firm Owns Four Different Software Lists
Ask a partner which software the firm uses and you will get a confident answer. Ask the office administrator, bookkeeper, senior paralegal, and newest associate and you will get four more. The subscriptions are real. The shared picture is not.
Start by separating four questions that firms routinely blend together:
1. What does the firm pay for?
This is the contract and invoice list: product name, plan, license count, price, renewal date, administrator, add-ons, and any bundled features. “Microsoft,” “Google,” or “Lexis” is not a usable entry. Record the exact plan and add-ons. Product families are neighborhoods, not addresses.
2. What is enabled?
A paid feature may be disabled by the administrator. An unpaid trial may be active. A vendor may add an AI button during an update. A browser extension may connect to systems the firm never intended it to reach. Enabled status must be checked in the actual account.
3. What is approved?
Approval means the firm has examined the exact feature, account, permitted information, users, settings, task, review, record, and limits. A familiar vendor name is not approval. A contract signed before the feature existed is not approval either.
4. What is actually used?
Usage logs, staff interviews, browser extensions, personal subscriptions, saved prompts, mobile apps, and workarounds reveal the operational list. This is where the firm usually finds both value and risk. The approved tool may sit idle while the unapproved tool has become part of Tuesday.
Paid is not approved. Enabled is not adopted. Familiar is not safe.
Each word answers a different question. Treating them as synonyms is how firms accumulate expense without control.
Run a 45-Minute License and Feature Inventory
Do not begin on vendor websites. Those pages describe everything the vendor would like to sell. Begin with invoices, renewal notices, administrator portals, accounting records, and the people who handle onboarding and departures.
Record these fields for every system
| Field | What to record |
|---|---|
| Identity | Product, exact plan, add-ons, contract entity, and business or personal account. |
| Ownership | Business owner, technical administrator, billing contact, and vendor contact. |
| Economics | License count, active users, monthly or annual cost, term, and renewal date. |
| Data | Information stored or reachable, system of record, connected services, and export path. |
| AI status | Available features, enabled status, eligible users, limits, and extra charges. |
| Decision | Approved uses, prohibited uses, reviewer, restrictions, and next review date. |
| Evidence | Current terms, privacy and security documents, written vendor answers, and usage reports. |
Check every software neighborhood
- Email, documents, spreadsheets, presentations, meetings, calendars, and file storage.
- Practice management, matter management, docketing, intake, CRM, forms, and client portals.
- Legal research, drafting, document analysis, citation tools, and knowledge libraries.
- PDF, e-signature, scanning, transcription, telephone, voicemail, and video systems.
- Billing, accounting, payments, timekeeping, expense, payroll, and reporting systems.
- Automation platforms, browser extensions, mobile apps, plug-ins, and connectors.
- Individual accounts that lawyers or staff use for firm work, whether reimbursed or not.
The last category is the awkward one. Include it anyway. An incomplete inventory protects no one, but it can look remarkably tidy in a meeting.
Translate AI Features Into Ordinary Work
Vendors describe AI with brand names. Firms should describe it with verbs. The label may change every quarter. The underlying capability usually performs one of a small number of jobs.
| Capability | Plain-English job | Possible firm use |
|---|---|---|
| Draft | Produces a first version from instructions or approved source material. | Email, letter, agenda, checklist, summary, or internal procedure. |
| Summarize | Condenses a document, conversation, or approved record. | Matter status, meeting recap, document overview, or client-update input. |
| Extract | Pulls named facts or fields into a structure. | Dates, parties, amounts, duties, deadlines, missing items, or document lists. |
| Compare | Identifies differences across versions or sources. | Contract clauses, draft changes, policy versions, or conflicting records. |
| Classify | Groups or labels items using defined criteria. | Intake routing, document type, issue category, or follow-up status. |
| Retrieve | Finds material in connected or selected sources. | Matter documents, prior work product, email, notes, or legal authorities. |
| Transcribe | Converts speech into text and may identify action items. | Internal meeting notes, approved interviews, or task lists. |
| Route | Moves information or triggers a later step. | Create a task, update a field, request review, or send an internal alert. |
A feature may perform several jobs. That does not make it a workflow. A workflow still defines the trigger, approved input, steps, owner, output, review, exceptions, record, and measure. The AI feature may assist with one step. It does not get to run the firm because it acquired a colorful button.
Match the Feature to a Workflow Problem
Feature lists invite the wrong comparison: which product has more checkmarks? Every vendor wins its own feature table. The firm needs a different test: can this exact feature improve a defined process under conditions the firm can control? The Small Firm Workflow Playbook lists 20 workflows to match against.
The seven-part fit test
| Fit question | What the firm must answer |
|---|---|
| Problem match | Does the capability address a recurring bottleneck the firm has measured? |
| Data proximity | Can it work where the approved information already lives without careless copying or broad access? |
| Output fit | Can it produce the format the next person or system actually needs? |
| Review cost | Can a qualified reviewer check the result faster than creating it through the current process? |
| Control | Can the firm limit users, data, permissions, sharing, connectors, and automated actions? |
| Adoption | Can the people who perform the work use it with a short instruction and a clear manual path? |
| Evidence | Can the firm measure time, corrections, rework, service, and exceptions against a baseline? |
Score each item 0, 1, or 2: no, uncertain, or yes. A low score does not mean the product is bad. It means the fit is bad. The distinction saves money and bruised egos.
Need the workflow side of the analysis? Use A Better Tuesday: One Small Workflow at a Time. It shows how to choose the bottleneck, define review, test in stages, and measure the full process.
Six Uses to Look for in Current Software
These examples describe categories, not endorsements. Availability, pricing, data protection, and controls differ by vendor, product, plan, configuration, and date. Check the exact account before using client information.
1. Email or document suite: client-update draft
Use an approved matter summary, not an unrestricted mailbox search, to draft a short update covering what happened, what comes next, what the client must do, and when the firm will communicate again. The lawyer verifies every fact, judgment call, commitment, and deadline before sending.
2. Meeting platform: notes to assigned tasks
Convert an approved meeting transcript or notes into a draft task list with owner, due date, dependency, and unresolved question. A participant checks the list before anything is assigned. Follow firm policy and applicable consent requirements before recording or transcribing.
3. Practice-management system: weekly matter report
Turn approved matter fields and status inputs into a standard internal report showing activity, next deadline, owner, open question, client contact, billing issue, and risk needing attention. Matter owners confirm the report. The system does not invent missing status.
4. PDF or document tool: fact extraction table
Extract named dates, people, amounts, document types, or obligations from a selected test set into a review table with source references. The reviewer checks the table against the original files. This is assembly assistance, not a substitute for reading what matters.
5. Legal research or drafting platform: issue map
Use approved facts to create a preliminary issue list, research plan, or first-draft outline. The lawyer researches current authority in authoritative sources, checks every citation and quotation, and decides what analysis belongs in the work product.
6. Intake or CRM system: missing-information list
Turn a completed fictional intake form into a structured list of people, dates, documents, stated goals, missing information, and questions for a lawyer. Keep conflict checking, matter acceptance, legal advice, and final communication in the firm’s approved process.
For more examples across intake, document collection, matter management, drafting, client communication, calendar, and billing, use the Small-Firm Workflow Playbook.
Paid Does Not Mean Approved
The phrase “we already pay for it” answers a budget question. It does not answer a confidentiality, competence, supervision, security, accuracy, or workflow question. The Vendor AI and Data Governance Audit is the review to run before approval.
Florida Bar Ethics Opinion 24-1 permits lawyers to use generative AI while keeping existing professional duties in place. Before confidential information enters a third-party AI system, the firm should analyze the exact use, terms, retention, sharing, data reuse, access, security, and any need for client disclosure or informed consent. The lawyer remains responsible for the work and professional judgment.
An AI feature inside a familiar suite may offer business controls that a personal account does not. It may also reach email, files, meetings, calendars, contacts, or connected applications. Familiarity reduces neither the need for review nor the size of the possible data path.
Seven checks before approving an existing feature
- Identity: What exact product, plan, account type, feature, version, and administrator are involved?
- Purpose: What defined task may the feature perform, and what tasks remain prohibited?
- Information: What may users enter, what can the feature retrieve, and what must remain outside?
- Terms: What do current terms say about retention, access, data reuse, training, ownership, deletion, and notice of changes?
- Controls: Can the firm limit users, permissions, sharing, connectors, automated actions, and external communication?
- Review: Who checks the result, against which source, before which use, and what triggers a hard stop?
- Record: What input, output, correction, approval, and decision evidence must be saved, and where?
Use the companion pillar Use AI Without Betting the Firm for Florida confidentiality, consent, supervision, billing, verification, advertising, and incident controls.
Approve the use, not merely the logo.
A vendor may sell several products under one brand. Plans, settings, protections, and data access can differ. Record the exact feature and workflow the firm approved.
The Hidden Risk Is Often the Connection
A stand-alone drafting box has a visible input. A connected assistant may have access to email, cloud files, calendars, meetings, contacts, matter data, shared drives, or other services. That access can make the feature useful. It can also make a small instruction produce a large data search.
The question is not merely, “Does the vendor train on our data?” Ask what the feature can retrieve, which permissions it inherits, whether users can connect outside services, what is logged, where prompts and outputs are retained, who can search them, and what happens when an employee leaves.
Inspect these connection points
- Shared mailboxes, group accounts, calendars, meeting recordings, and transcripts.
- Firm-wide drives, broad folders, inherited links, external shares, and old client workspaces.
- Practice-management, CRM, accounting, payment, and document-management connectors.
- Browser extensions and plug-ins that can read the active page, document, or clipboard.
- Mobile apps, personal devices, downloads, copied text, screenshots, and local caches.
- Automated sends, task creation, field updates, client messages, and other actions that bypass review.
Do not turn on a broad connector to solve a narrow task.
Start with the least information, access, and automation needed for the approved workflow. Convenience is not a permission model.
Use a One-Page Current-Stack Audit
The inventory becomes useful when it produces a decision. For each candidate feature, record one page that a partner, administrator, and reviewer can understand without decoding the vendor’s product map.
| Audit field | Decision to record |
|---|---|
| Problem | What recurring work or delay are we trying to improve? |
| Current process | Who does the work now, using which sources and systems, in how much total time? |
| Candidate feature | What exact product, plan, feature, account, settings, and users will be tested? |
| Approved information | What fictional, public, or otherwise approved input may be used in the first test? |
| Expected result | What output should the feature produce, in what format, for which next step? |
| Controls | What permissions, connectors, sharing, retention, and automated actions are limited or disabled? |
| Review | Who checks the result against which source, and what requires rejection or escalation? |
| Measure | Which time, correction, rework, service, adoption, and cost measures decide the result? |
| Decision date | When will the firm approve, revise, reject, replace, or buy? |
If the page cannot name the problem, account, data, reviewer, and measure, the firm is not ready to test the feature. It is ready to watch another demonstration.
Make the Buy-or-Do-Not-Buy Decision
“Use what we already own” is a disciplined starting point, not a religion. Current software may be the wrong place for a task. A specialized legal product may provide better sources, matter boundaries, review tools, integrations, or controls. The firm should buy when the gap is real and the proof is better than the pitch.
A new tool may be justified when
- The current stack lacks the capability required by a frequent, valuable workflow.
- The existing feature produces too many material errors or requires review that consumes the savings.
- The current data boundary, permissions, retention, or contract terms do not fit the proposed use.
- A specialized product provides authoritative sources, matter-level controls, citations, or review functions the task requires.
- The new product removes manual transfers or duplicate entry that create delay and error.
- A controlled test shows a net gain after production, review, corrections, training, administration, and subscription cost.
- The firm has an owner, rollout plan, user group, support path, and date for deciding whether the product stays.
Evidence to demand before purchase
- A written workflow problem and baseline, not a list of admired features.
- A test using representative, approved material, including difficult and flawed examples.
- Written answers on data, terms, security, permissions, connectors, retention, deletion, and incident response.
- A comparison with the current process and current software, using the same starting and ending points.
- Named users, reviewer, administrator, training, support, renewal owner, and cancellation date.
- A decision rule that allows the firm to say no after the trial.
Use the AI Vendor Evaluation Worksheet and the Vendor AI & Data Governance Audit when the current stack cannot do the job and a new product reaches the short list.
Cut the Tool Cemetery
An inventory should not merely find more features. It should find expense that no longer earns a place. Small firms often pay twice for transcription, scheduling, storage, e-signature, PDF functions, task management, CRM, chat, automation, or AI drafting because each purchase solved one urgent afternoon.
Cancellation is not the goal. A cleaner operating system is. Keep a product when it supports an approved workflow, fits the data and review requirements, has an owner, and produces enough value to justify its cost and administration. Retire it when no one uses it, another approved system performs the same job, or the workflow disappeared months ago.
Measures that expose the truth
| Measure | What it reveals |
|---|---|
| Inventory coverage | Percentage of paid systems with exact plan, owner, renewal, users, data, and AI status recorded. |
| Approval coverage | Percentage of enabled AI features with a written decision to approve, restrict, disable, or investigate. |
| Workflow use | Number of current features tied to a defined, trained, measured workflow. |
| Adoption | Eligible users who perform the approved workflow correctly during the review period. |
| Net value | Time, quality, service, or cycle-time gain after review, corrections, administration, and cost. |
| Waste removed | Unused licenses, overlapping tools, personal subscriptions, or add-ons retired or consolidated. |
| Control failures | Exceptions, prohibited uses, data incidents, missed reviews, or automated actions outside scope. |
License utilization is not enough. A person can click an AI button every day and create no value. Measure whether the approved workflow improves the finished result. Activity is what software companies report when outcomes are feeling shy.
Seven resources. One sensible starting point.
You will receive the kit immediately, plus the Small Law Firm AI Workflow Brief, practical emails about useful workflow changes, Florida guardrails, and tool updates that matter to firms under 25 lawyers. Unsubscribe anytime.
- Firm AI-readiness checklist
- AI tool-approval scorecard
- Five workflows worth testing first
- Daily lawyer verification checklist
- Plain-English guide to Florida Bar Opinion 24-1
- Sample client language and incident-response protocol
- 90-day implementation outline
A 14-Day Current-Stack Audit
Two weeks is enough to replace assumptions with a working list and one test decision. It is not enough to inspect every feature in every product. Good. The firm needs a repeatable audit method, not a technology census conducted by candlelight.
Days 1–2: Build the paid-software list
Collect invoices, renewal notices, contracts, administrator portals, and expense reports. Record exact products, plans, add-ons, licenses, active users, owners, costs, and renewal dates. Add personal and reimbursed accounts used for firm work. Mark missing owners, inaccessible accounts, former-employee access, and unknown renewals for action.
Days 3–5: Find enabled and available AI features
Review current administrator and vendor documentation for the exact plan. Record whether each feature is included, add-on, trial, metered, disabled, or limited by role. Identify the data each feature can receive, retrieve, create, change, or send. Interview representative users about actual use, workarounds, browser extensions, and personal accounts.
Days 6–8: Match one feature to one problem
Choose one frequent, painful, repeatable, reviewable, recoverable, and measurable task. Apply the seven-part fit test and define the current baseline. Write the expected output, reviewer checks, hard stops, and manual path. Reject candidates that require broad data access to solve a narrow problem.
Days 9–11: Check the feature and test safely
Review the exact plan, terms, settings, permissions, data path, connectors, retention, and controls. Use fictional, public, or otherwise approved material with ordinary, ambiguous, incomplete, and flawed examples. Record production time, review time, material corrections, omissions, and user effort. Stop if the feature reaches unapproved information, bypasses review, or changes the task beyond scope.
Days 12–14: Decide and clean up
Approve, restrict, disable, revise, or reject the feature for the defined use. Record the decision, owner, approved users, instructions, review, restrictions, and next review date. Identify duplicate or unused licenses for cancellation, consolidation, or renewal negotiation. Create a short list of gaps that might justify a new product and the evidence required before purchase.
The audit should end with a decision and an owner. “We should look into that” is neither. It is where software costs go to reproduce.
Frequently Asked Questions
How do we know whether our current software already includes AI?
Check the exact product and plan in the administrator portal, current contract, invoice, and official vendor documentation. Then check whether the feature is enabled for the firm and for which users. A vendor’s general product page may show features your plan does not include.
If an AI feature is included, is it safe to use with client information?
No automatic conclusion follows from “included.” Review the exact account, terms, settings, permissions, retention, data reuse, access, connectors, task, users, review, and records. Analyze Florida confidentiality duties and any need for disclosure or informed consent before confidential information enters a third-party AI system.
What if a vendor enables an AI feature automatically?
Treat the change as a review trigger. Determine what was enabled, who can use it, what information it can reach, which defaults apply, and whether it can create or send anything. Disable or restrict the feature until the firm makes a documented decision when the answers are unclear.
Should we choose Microsoft 365, Google Workspace, or another suite based on AI?
Usually not by feature count alone. Start with the firm’s current environment, workflows, data, user habits, controls, support, and switching cost. Compare the exact plans against a defined task. A feature that looks better in a demonstration may be worse inside the firm’s actual process.
Should every lawyer receive the same AI licenses?
Not necessarily. Match licenses to approved workflows, roles, volume, review responsibilities, and training. A small pilot group often produces better evidence than a firm-wide purchase. Expand after the workflow, controls, and value are stable.
Is a vendor promise not to train on our data enough?
No. Training is one issue. Also examine retention, human and administrator access, logging, data location, sharing, subprocessors, connectors, permissions, deletion, security, breach response, ownership, and changes to terms. Then decide whether the exact workflow fits.
How should we compare a general business AI feature with a legal-specific product?
Compare both against the same workflow and baseline. Examine source authority, matter boundaries, legal research and citation functions, data controls, integration, output quality, reviewer effort, training, administration, and total cost. The legal label does not end the analysis; neither does the lower price.
When should we buy a new AI product?
Buy after a measured workflow need survives the current-stack audit and a controlled test shows that the new product produces a net gain with acceptable controls. Name the users, owner, reviewer, support path, renewal date, and stop rule before signing.
How often should the firm repeat the audit?
Review the inventory at least annually and before material renewals. Also review after a major feature, plan, term, data-protection, connector, administrator, incident, or workflow change. Product updates do not wait for the annual partner retreat.
Bring the firm’s software list and one recurring task that wastes time.
We will identify what the firm may already own, which feature is worth checking, what information is involved, where lawyer review belongs, and whether the right answer is use it, restrict it, replace it, or buy nothing.
No sales presentation. No product quota. No suggestion that every problem requires a new subscription.
About Greg Yates
Greg Yates is the founder of Actionable AI and a former Big Law corporate equity partner at Steptoe (NYC Managing Partner) and Seyfarth Shaw. He helps Florida law firms with fewer than 25 lawyers improve recurring work using AI capabilities in software they may already own. He is independent, does not sell software, and will say when a product or AI itself is not the answer.
Learn more about Greg, browse the Florida AI Guides, or review Actionable AI services.
Primary Authorities and Current Vendor Documentation
- Florida Bar Ethics Opinion 24-1: Generative AI, confidentiality, oversight, fees, and advertising.
- Rules Regulating The Florida Bar: Current professional-conduct rules and updates.
- The Florida Bar Guide to Getting Started with AI: Florida Bar educational guidance on cautious experimentation and review.
- ABA Formal Opinion 512: Additional guidance on competence, confidentiality, communication, supervision, candor, and fees.
- NIST AI Risk Management Framework: Voluntary framework for governing, mapping, measuring, and managing AI risk.
- Microsoft 365 Copilot Plans and Pricing: Current Microsoft plan and add-on information; availability and terms change.
- Google Workspace with Gemini: Current Google Workspace feature and access guidance; availability depends on plan and configuration.