When to Outsource Tax Research vs Keep It In-House
Most firms with fewer than five practitioners should keep tax research in-house, using AI-assisted tools to handle volume without adding headcount. Firms above ten practitioners benefit from a hybrid model: a dedicated researcher for recurring complex work, with AI tools covering the 80% of questions that don't need that level of expertise. Outsourced research services fill a specific gap for seasonal surge capacity and one-off specialty questions, but they're rarely cost-effective as a primary research method. The deciding factors are your question volume, your average question complexity, and whether faster research translates directly into more billable capacity at your firm size.
Three Ways to Handle Research
Every firm picks from the same three options, whether they think about it explicitly or not.
Option 1: Dedicated research staff. You hire someone whose primary job is answering research questions for the team. They use your databases, develop institutional knowledge about your clients, and deliver consistent quality. The downside is the fixed cost: salary, benefits, desk space, and software licenses, whether the work fills their week or not.
Option 2: Outsourced research services. You send questions to a contract CPA, an EA, or a specialized research firm, paying per question or on retainer. Zero fixed cost and access to specialists you couldn't afford full-time. The downside is turnaround time (usually 24 to 72 hours), loss of client context, and the need to review every work product before relying on it.
Option 3: AI-assisted in-house research. Each practitioner handles their own questions using purpose-built tools. Questions that would take 30 minutes in a traditional database take 3 to 5 minutes. No new headcount, no outsourcing lag. The practitioner still does the work, but at a fraction of the time cost.
Most firms default into Option 3 without realizing it, because "I'll just look it up" is the path of least resistance. The question is whether that default still makes sense as your firm grows, or whether another option starts winning on economics.
The Cost Math at Three Firm Sizes
The real numbers depend on your billing rate, question volume, and local labor market. But the ratios hold across most practices. I'm using a $300/hour blended billing rate, which is a reasonable midpoint for experienced practitioners doing individual and small business work.
If you want to see how the cost of tax research per return breaks down at the per-question level before reading further, that post covers the underlying unit economics.
Solo Practitioner (1 CPA)
A solo practitioner averaging 2 research questions per day during a 16-week filing season handles roughly 160 questions.
| Approach | Cost | Annual total |
|---|---|---|
| Traditional database (30 min/question) | 80 hrs opportunity cost + subscription | $27,000-30,000 |
| AI-assisted (5 min/question) | 13 hrs opportunity cost + subscription | ~$5,000 |
| Outsourced ($75-150/question) | 160 questions at per-question fee | $12,000-24,000 |
| Dedicated researcher | Not viable at this scale | $55,000-75,000 |
At one practitioner, the dedicated researcher is out. You can't justify a $65,000 salary for 160 seasonal questions plus whatever comes up the rest of the year. Outsourcing at $75 to $150 per question gets expensive fast and introduces a delay that kills the "client is on the phone right now" scenario.
AI-assisted self-service wins on every dimension for solo practitioners. The total cost (tool subscription plus your time) is under $5,000. Everything else is $12,000 or more.
3-Person Firm
Three practitioners at 2 questions per day each generate roughly 480 research questions during filing season, with maybe another 200 scattered across the rest of the year. Call it 680 annual questions.
| Approach | Cost | Annual total |
|---|---|---|
| Traditional database (30 min/question) | 340 hrs opportunity cost + subscription | $107,000-114,000 |
| AI-assisted (5 min/question) | 57 hrs opportunity cost + subscription | ~$19,400 |
| Outsourced ($75-150/question) | 680 questions at per-question fee | $51,000-102,000 |
| Dedicated researcher (full-time) | Salary + benefits + tools | $70,000-95,000 |
This is where it gets interesting. A dedicated researcher at $80,000 all-in starts to compete with 340 hours of practitioner time. But the researcher doesn't eliminate the time cost entirely. Practitioners still need to formulate the question, wait for the answer, and review the work product. Realistically, the researcher handles the deep dives (the 90-minute questions), while practitioners handle simple lookups themselves.
The hybrid that works at this size: AI-assisted tools for every practitioner, covering the 70% of questions that are straightforward lookups and single-authority answers. Save outsourcing for the occasional multi-jurisdiction or novel issue that would take anyone more than an hour. You'll outsource maybe 30 to 50 questions per year at $100 to $150 each. Total research cost: roughly $22,000, versus $107,000 using traditional databases alone.
Building a research playbook is what makes this hybrid model work at the 3-person stage. Without clear protocols, each person reinvents their research process on every question.
10-Person Firm
Ten practitioners generating research questions across individual, business, trust, and estate work produce 2,000 or more questions per year. Complexity skews higher because larger firms take on more complex engagements.
| Approach | Cost | Annual total |
|---|---|---|
| Traditional database (30 min/question) | 1,000+ hrs opportunity cost + subscription | $305,000-312,000 |
| AI-assisted (5 min/question) | 167 hrs opportunity cost + subscription | ~$55,000 |
| Dedicated researcher (full-time) | Salary + benefits + tools | $70,000-95,000 |
| Outsourced (100 complex questions) | Per-question fee | $15,000-25,000 |
At ten practitioners, the dedicated researcher pays for itself. One person at $85,000 all-in handles the deep research (multi-authority, multi-jurisdiction problems consuming 60 to 90 minutes each), freeing practitioner time worth $300/hour. The break-even is roughly 280 practitioner hours saved per year, less than 6 hours per week.
The optimal model: a dedicated researcher for the complex 20%, AI tools for every practitioner on the routine 80%, and outsourced specialists for the 2 to 3% of questions in areas the firm rarely sees (international tax, ERISA, tax-exempt organizations). Total: roughly $120,000 to $140,000 per year, versus $300,000-plus in practitioner time under the traditional approach.
The Break-Even on Hiring a Full-Time Researcher
The math comes down to one ratio: how many practitioner hours does this person displace, and what are those hours worth?
A competent tax researcher handles 8 to 12 complex questions per day. At 10 questions per day across 230 working days, with an average of 40 minutes saved per question, that's 1,533 practitioner hours freed per year. At $300/hour, those freed hours represent $460,000 in potential billing capacity against an all-in cost of $80,000 to $95,000.
But freed hours only generate revenue if you fill them with billable work. If you don't have additional clients to serve, the savings are real but they show up as capacity, not cash. And the researcher needs a consistent pipeline. If your question volume drops below 6 to 8 complex questions per day, you're paying for idle time. That's why the break-even sits around 8 to 10 practitioners.
The Regulatory Floor on Research Quality
Regardless of who does the research, the quality standard is the same. Treas. Reg. 1.6694-1 defines the due diligence requirements for paid preparers. If you outsource research, you're still the signing preparer, and you're still responsible for the position taken on the return.
Treas. Reg. 1.6662-4 sets the other boundary. A substantial understatement penalty applies unless the position has "substantial authority," which requires actual legal analysis from primary sources, not a vendor's summary email. The IRS Paid Preparer Due Diligence requirements reinforce this: your obligation doesn't transfer just because someone else looked up the answer.
This is why outsourcing works only when you review the work product yourself. The outside provider gives you a starting point. You verify citations, confirm the analysis applies to your client's facts, and make the call. Skip that review step, and you're taking on risk that no engagement letter covers.
The billable-hour economics of that review step matter, too. If outsourced research saves you 30 minutes of lookup time but adds 15 minutes of review time, your net savings is 15 minutes, not 30.
What Changes the Calculus
Seasonal compression. Most firms have two peak windows: January through April for original filings and September through October for extensions. If 80% of your research volume concentrates in those 24 weeks, a full-time researcher is underutilized for the other six months. AI tools and outsourcing handle those surges without carrying fixed cost year-round. Firms with steady off-season work (estate planning, advisory, IRS representation) can justify the full-time hire more easily because the research pipeline doesn't dry up in summer.
Question complexity distribution. Firms that generate mostly simple lookups get maximum value from AI tools. Firms with primarily interpretive, multi-source questions hit the dedicated researcher break-even sooner.
Staff experience levels. Newer practitioners generate more research questions per return. A growing firm will see research volume spike before it stabilizes. Building a research playbook for your first staff accountant is the difference between that spike costing you $50,000 in lost practitioner time and costing you $5,000 in tool subscriptions.
Practice area mix. When comparing traditional databases to AI tools, editorial databases still have an edge on narrow specialties like international tax treaty interpretation or ERISA compliance. AI tools cover the breadth of domestic federal and state tax law that constitutes 90% of most firms' research needs.
The Decision Framework
Here is the simplified version.
Under 5 practitioners: Keep research in-house. Give every practitioner AI-assisted research tools. Outsource only for genuine specialty questions you see fewer than 5 times per year. Total research budget: $5,000 to $15,000, depending on your tool choices and question volume.
5 to 10 practitioners: Hybrid. AI tools for every practitioner. Budget $3,000 to $5,000 per year for outsourced specialty research. Consider a part-time researcher (20 hours/week) if your complex question volume exceeds 8 per day consistently. Total: $30,000 to $60,000.
Above 10 practitioners: Full hybrid. Dedicated researcher for complex work. AI tools for every practitioner. Outsourced specialists for niche practice areas. The researcher pays for themselves in freed practitioner hours within the first quarter. Total: $100,000 to $150,000, saving $150,000-plus versus the traditional database approach at scale.
Run the Numbers on Your Own Firm
The framework above gives you the structure. The specific numbers depend on your billing rate, question volume, and staff mix. Track your research questions for two weeks. Count them. Categorize them: simple lookup, moderate (single authority, under 20 minutes), and complex (multiple authorities, 30 minutes or more). Multiply each category by your cost per question.
That exercise will tell you more than any vendor comparison. We built Tax Orator to handle the 80% of research questions that don't require a specialist, because that's where the time goes. Whether it's the right tool for your firm depends on what your numbers look like. See the pricing page and run the math.