A client sends you rough notes on a Tuesday evening and asks for a finished audio episode. You smooth the text so it sounds natural read aloud, generate it in a cloned voice, fix the pacing and lay a little music under the intro. Done the traditional way, that one episode would have cost them between $500 and $2,000.
Now look at your own week. A job that pays the same as it did three years ago while groceries do not. A manager who calls you a valued team member and then freezes hiring. Maybe a side gig that pays in exposure and late invoices. And everywhere, people saying AI will flatten work like yours.
Here is the twist. The same tools that scare everyone are the ones you would be selling. There are over 50 million content creators worldwide, and most of them cannot pay for professional voice talent. They need someone who already knows the tools, handles consent properly and delivers clean audio every week. The agencies that set up now become that someone before the creators even start shopping around. Later arrivals compete on price with everyone who waited.
Setup runs anywhere from $0 to $300. The two standard retainers are $1,000 a month for basic audio conversion and $2,000 a month for full production, and two to four weeks is a normal run-up before your first paid month.
Tonight, pick one blog post from a company whose writing you already follow and rewrite its opening paragraph the way a person would actually say it out loud. That small rewrite is the first sample in your portfolio.
You have probably noticed it yourself. Everyone you follow is making more video than they did a year ago, and almost none of them can afford a voice actor to go with it. Companies want one consistent brand voice across thousands of touchpoints, from the phone menu you sit through to their training videos to their customer service bots. Podcasts keep multiplying, and millions of shows need someone to make them sound professional. People are listening to more audiobooks than ever, while the old way of producing one still prices most authors out. All of those pressures are landing at once, and that is why this field is growing so fast.
Think about what your future clients pay today. Traditional voiceover studios charge $500-2,000 per day, and professional voice actors command $300-1,000 per hour. With AI voice synthesis you can deliver comparable quality at roughly one-tenth the cost, and your turnaround is measured in minutes instead of weeks. Which of the creators you already watch would jump at that?
Before you build anything, you need an honest picture of what today's AI voices can deliver and where they still fall short. The field has moved fast. Knowing your limits is what keeps your clients happy, because nobody forgives a promise you could not keep.
So you know the demand is there and roughly what the tools can do. The next thing a client will ask you is simple: what exactly would they be buying from you?
This is the highest-demand service in voice synthesis right now. YouTubers with established English-language audiences want to reach Spanish, Portuguese, Hindi, German and other big language markets without losing the personal connection their viewers feel with them.
The podcast market has matured, but production costs still stop a lot of people from starting one. You can serve newsletter writers, authors, executives and thought leaders who want a podcast without giving it their time.
Big companies spend millions on how they look and often forget how they sound. You can build them a custom AI voice they "own" and use consistently everywhere audio shows up.
The audiobook market is worth more than $5.3 billion and grows 25% a year. Traditional production costs $5,000-50,000, which prices out most independent authors. With AI voice, you can give those authors an audiobook they could never have afforded.
Radio spots, podcast ads, YouTube pre-roll and social ads all need a voice. With AI you can give an advertiser dozens of variations to A/B test, at a fraction of what a studio would charge.
Online education is a huge market, and production costs hold a lot of course creators back. You can take their written course material and turn it into professionally narrated modules.
One $1,000 basic audio retainer is enough to change how your month feels. After tools and production costs, it could pay the car insurance and your internet bill, all from turning one client's blog posts into narrated episodes. Which service on this menu could you deliver well next week? Pick that one and make a sample a client can actually hear.
The tools you choose decide what you can offer and how you stand against everyone else doing this.
You do not need all of these on day one. If you had to pick just two, which would you start with? For most people, ElevenLabs plus Descript covers the first few months.
Make your demo content:
- Dubbed video demo: Take a 2-minute clip from a public domain video or your own content and dub it into Spanish and German
- Podcast demo: Turn a popular newsletter edition into a 10-minute podcast episode
- Corporate demo: Create a professional IVR greeting and a short training module
- Advertising demo: Produce the same 30-second ad in 5 different styles
Build your presence:
- A simple portfolio website (Carrd, Webflow, or Squarespace)
- A LinkedIn profile set up for voice synthesis services
- Case studies written up from your demo projects (treat them as real)
- A pricing page or PDF with clear service tiers
Weeks 5-6: Outreach Campaign
Build your prospect list:
- YouTubers in your target subscriber range with English-only content
- Newsletter writers with 10K+ subscribers who could podcast
- Corporate training departments at mid-size companies
- Indie authors with upcoming book releases
How you reach out:
- Personalized Loom videos showing their own content dubbed
- Direct email with a specific reason this helps them
- LinkedIn connection requests that offer something first
- Free 30-second samples made from their actual content
Your daily targets:
- 10 personalized outreach messages
- 5 follow-ups on earlier outreach
- 2 sample creations for high-value prospects
Be honest with yourself here: could you hold 10 messages a day for two weeks while your day job is busy? If not, cut the number now and keep it steady. A smaller daily habit you actually keep will carry you further than a big plan you drop by Thursday.
Weeks 7-8: First Client Delivery
Turning interest into a sale:
- Offer pilot projects at reduced rates in exchange for case study rights
- A standard first project: $500-1,000 for proof of concept
- Document everything for future case studies
Delivering well:
- Over-communicate throughout the project
- Build a revision buffer into your timelines
- Gather detailed feedback so you get better
- Ask for a testimonial right after a successful delivery
Imagine hearing your studio name in the credits of a client's podcast intro, in a voice you produced. That is something you can point at for years. The $500 proof of concept is what gets you there, and the day that first testimonial lands is the day it becomes your agency. Ask for it the hour you deliver.
Setting your prices
You know what you can sell. Putting a number on it is a separate decision, and it is the one that decides whether $2,000 a month is your floor or your ceiling.
Value-Based Pricing Principles
Price on the value you deliver, and treat your costs as the floor underneath that. A dubbed video that opens a creator up to 100 million potential Spanish-speaking viewers is worth far more to them than the $50 in tool costs it took you.
Work out what it is worth to your client: For a YouTuber with 1M subscribers earning $3-5 CPM, each new language channel could bring in $3,000-5,000/month in extra revenue within a year. Next to that, your $2,000/month service is a small line in their budget. When a prospect pushes back on price, that is the comparison to walk them through.
Tier Structure
Starter Tier (Entry Point):
- Single video dubbing: $500
- Podcast pilot episode: $750
- Voice clone setup + test: $1,000
- Purpose: A low-risk trial for hesitant clients
Professional Tier (Core Business):
- Per-minute dubbing: $50
- Monthly podcast production: $2,500
- Corporate voice development: $10,000
- Purpose: Where most of your revenue comes from
Premium Tier (Whale Clients):
- Global launch package (5 languages): $15,000
- Full voice identity system: $25,000
- Annual enterprise contracts: $50,000+
- Purpose: These change your business when you land one
Retainer vs. Project Pricing
Retainers give you predictable income and closer relationships with your clients. Push for them whenever you can by offering:
- Priority turnaround (24-48 hours vs. 1 week)
- Rollover minutes/projects
- Dedicated account management
- Volume discounts built in
What would one steady $2,000 retainer do for the way you sleep, compared with chasing a new $500 job every week? For most people, the retainer is the moment this starts to feel like a business.
Picture two retainers landing on the first of the month. That is rent covered before you open your laptop, and the overtime shifts you used to beg for become optional. Each client who says yes is one less night spent worrying about whether the car will pass inspection.
Protecting yourself legally
Voice synthesis sits in a legal grey area, and the rules are changing quickly. Protect yourself and your clients with proper paperwork and careful habits.
Required Contracts and Permissions
Voice owner consent: NEVER clone a voice without explicit written permission from the person it belongs to. Your contract must cover:
- Scope of permitted use (platforms, content types, geographic regions)
- License duration (perpetual, annual, project-specific)
- Exclusivity terms (can you use this voice for other clients?)
- Revision and update rights
- Ownership of the voice model (who controls it, who can delete it?)
Client service agreements: A standard MSA covering:
- Deliverables and specifications
- Revision limits (3 rounds is standard)
- Payment terms (50% upfront, 50% on delivery)
- Intellectual property ownership
- Confidentiality provisions
- Limitation of liability
This is a legal minefield, and I would stay well clear of it. If you use the recognizable voice of a celebrity, politician or public figure without an explicit licensing agreement, you open yourself up to right of publicity claims, trademark infringement and fraud allegations. Do not offer celebrity voice cloning unless you have proper licensing agreements in place.
Regulatory considerations
Three sets of rules matter to you, and which ones apply depends on what the voice is used for rather than on how you made it.
Disclosure and synthetic media. The EU AI Act carries transparency obligations for synthetic audio and video, and several US states have passed laws on deepfakes and voice cloning, with California, Texas and New York among the most active. For you as an agency, disclosure is becoming the default expectation for synthetic voice in anything the public hears. Building it into your deliverables now is cheaper than adding it later.
Where voice work crosses into regulated calling. Most agencies in this space have no idea this boundary exists, and it carries the biggest financial exposure you will face.
In February 2024 the Federal Communications Commission adopted a Declaratory Ruling confirming that the Telephone Consumer Protection Act's restrictions on an "artificial or prerecorded voice" cover AI technologies that generate human voices. The FCC's reasoning is that a cloned or synthesised voice is artificial because no person is speaking. So AI voice used in outbound calls to US consumers falls inside the TCPA, which requires prior express written consent for artificial-voice calls.
The TCPA carries statutory damages of $500 per call, rising to $1,500 for wilful or knowing violations, with no cap and a private right of action. Those figures count every single call, which is what turns this from expensive into business-ending. A modest outbound run of ten thousand calls without valid consent is a theoretical exposure in the millions. Picture that letter arriving, with your studio's name on the production.
None of that touches dubbing a YouTube video, producing an audiobook or building a corporate voice identity, which is most of the work in this guide. It applies the moment a client takes the voice you produced and puts it on an outbound dialler. That leaves you with two habits to build.
First, ask what the voice is for, in writing, before you build it. A client commissioning a "customer outreach voice" is describing something very different from a narrator, and the difference is a federal statute.
Second, say in your contract that the client is responsible for consent and compliance in how the output is deployed. You are producing an asset. The client is the one making the calls, and your contract should say so plainly instead of leaving it to be argued over after a complaint.
Call recording, if you build interactive agents. If your work grows into conversational agents that handle live calls, recording consent becomes a separate obligation, and in the US it varies by state. Several states require every party to consent to a recording. An agent that records by default is making a compliance decision for whoever deployed it, so make that default a deliberate choice and write it down with your client.
Growing your agency
Phase 1: Solo Operator ($5K-15K/month)
Client Load: 3-5 active clients Focus: Quality delivery, relationship building, case study development Time Allocation:
- 40% production work
- 30% client management
- 20% business development
- 10% admin
Phase 2: Small Team ($20K-50K/month)
When to hire:
- You are turning away work because you have no capacity
- Production time is eating the time you need for sales
- Quality is slipping because you are overloaded
First Hires:
- Freelance translator/localization specialist
- Audio editor for post-production
- Part-time project coordinator
Phase 3: Full Agency ($100K+/month)
Team Structure:
- Account managers for client relationships
- Production team (3-5 specialists)
- QA team for quality control
- Sales/business development
- Operations manager
Additional Revenue Streams:
- White-label services for marketing agencies
- Training and consulting for in-house teams
- Software/SaaS products built on your workflow
With three to five clients on retainer, the solo phase on this page sits at $5K to $15K a month. Hold near the top of that for a season, keep enough after costs to ride out a lost client, and you could tell your parents they can drop the Saturday shift, and mean it. Guard the 40% of your week that goes on production, because that is what keeps the retainers renewing. Do you actually want a team, or would you rather stay solo at the top of that range? Decide early, because the answer changes who you hire and when.
Mistakes to steer around
1. Overpromising Voice Quality
Some voices just do not clone well. Some people have acoustic qualities that current technology struggles to copy. Always run a test clone before you commit to a project.
Prevention: Make a voice quality assessment part of every cloning contract before anyone signs. Add clauses that let either side cancel if the clone does not meet the agreed quality.
2. Ignoring Translation Quality
A bad translation ruins content however good the voice is. Machine translation has improved, but it still makes errors a native speaker spots instantly.
Prevention: Always have a human review translations, even when AI does the first pass. Build translation costs into your prices.
3. Underpricing Your Services
You are selling premium technology that delivers real value. Racing to the bottom on price brings you bad clients and burns you out.
Prevention: Work out your true costs, including your own time. Add a healthy margin. Talk about price in terms of what the client gets back, and keep your costs to yourself.
4. No Revision Limits
Unlimited revisions sound friendly, and they quietly turn into scope creep and projects that lose you money.
Prevention: Standard contracts include 2-3 revision rounds. Anything beyond that is billed at your hourly rate.
5. Skipping Contracts
Working off verbal agreements or email chains is a risk you do not need. One dispute over voice rights can sink your business if the paperwork is missing.
Prevention: No work starts without a signed contract. Use lawyer-reviewed templates. Document every permission and approval.
6. Failing to Specialize
When you try to serve everyone, you end up serving nobody well. The agency that is "best at YouTube dubbing for fitness creators" wins more business than the agency that "does all voice stuff."
Prevention: Choose your niche based on where your first clients succeed. Build your expertise and reputation there. If you looked at your last three conversations with prospects, which niche keeps coming up?
How you will know it is working
Leading Indicators of Success
| Metric | Target |
|---|
| Response rate to outreach | 10%+ |
| Proposal-to-close ratio | 25%+ |
| Client satisfaction scores | 4.5+/5 |
| Referral rate | 30%+ of new clients from referrals |
Lagging Indicators
- Monthly recurring revenue growth
- Average revenue per client
- Client lifetime value
- Profit margins (target: 60%+)
If your outreach response rate sits well under 10% after a few hundred messages, what does that tell you? Usually it means your sample or your list needs work, and it says very little about your ability. Fix the sample first.
What could go wrong
Market Risks
Technology Commoditization: As voice synthesis gets easier, prices will come under pressure. Protect yourself with relationships and service quality that clients value beyond the tool itself.
Platform Changes: ElevenLabs, Rask.ai, or other core platforms could change their pricing, features, or policies. Stay comfortable on more than one platform.
Operational Risks
Key Person Dependency: If you are the only one who can do the work, you are capped and exposed. Write your processes down and train others.
Client Concentration: Relying on 1-2 clients for most of your income is dangerous. Spread your client base.
Financial Risks
Cash Flow: Project income is lumpy. Build a reserve and push for retainers.
Non-Payment: Some clients will not pay. Take deposits, use contracts, and have a collections process ready.
If your biggest client left next month, how many weeks could you cover your own bills? That number is your real safety margin, and it is worth knowing before it is tested.
When the money might arrive
You have the services and the prices. The question left is when any of it shows up in your account.
Month 1: $0-500 (learning, building portfolio, initial outreach) Month 2: $500-2,000 (first pilot projects, discounted rates) Month 3: $2,000-5,000 (first full-price clients, referrals starting) Month 4-6: $5,000-10,000 (established workflows, repeat clients) Month 7-12: $10,000-20,000 (multiple retainer clients, premium projects) Year 2: $20,000-50,000 (small team, specialized reputation)
These figures assume you keep putting in steady effort and that you do land clients. Your results will vary with the niche you pick, how well you sell, and what the market does. Please treat them as a range of what is possible, with no promise attached.
Where you go from here
The voice synthesis market is still early. The technology will keep getting better, prices will keep falling, and more businesses will adopt it. If you build expertise, relationships and efficient habits now, you will be in a strong position as the market grows.
Your first step is simple. Sign up for ElevenLabs, clone your own voice, and make something. Hear the technology for yourself. Then pick one potential client and make a sample from their content. The creators, companies and content producers who need you exist today. They are publishing in one language, paying too much for voiceovers, or avoiding audio altogether because of the cost. You can solve that for them while you build a profitable business of your own.
2026 Market Snapshot
In 2026 the voice synthesis agency sits where two Independent market research theses meet: the Voice Cloning report (creators building "audio content machines") and the Agencies report (AI-powered micro-agencies winning on speed and price). If you are working solo, your playbook is to package ElevenLabs, Descript and Resemble outputs into productized services for podcasters, course creators and localization buyers who have no wish to learn the tools themselves.
- Margin profile: AI-powered micro-agencies routinely run 50-70% gross margins on service work
- Documented case studies: Brett Williams (Designjoy) grew an adjacent productized agency to $1,500,000 ARR while building in public; Alex West publicly documents CyberLeads revenue
- Demand signal: AI video generation and editing is the fastest-growing skill in Upwork's In-Demand Skills 2026 report at +329%, and synthetic voice sits inside that same production pipeline
- Adoption gap: less than 5% of small and mid-market businesses have implemented meaningful AI automation
- Production economics: ElevenLabs voice clones typically cost $0.10-$0.30 per 1,000 characters, undercutting human voiceover by 80-95%
Key Players to Watch
These are publicly circulated figures, most of them self-reported by the person named. We have not independently confirmed them, and none is adjusted for costs, so read them with that in mind.
The 2026 list combines voice cloning platforms, the agencies wave, the buyers downstream of you, and educator-operators teaching the agency model.
- ElevenLabs: leading consumer-grade voice cloning platform, current default for agencies
- Descript: integrated podcast editor with realistic voice cloning
- Resemble AI: dynamic voice content for enterprise and games
- Respeecher: voice cloning for film, TV, and content
- Deepsync, Murf, Play.ht: additional providers in the operator's stack
- DeepZen, Veritone Voice Network: enterprise audiobook and localization platforms
- Flawless: AI dubbing platform for film localization
- Liam Ottley: YouTube educator who codified the AI-powered agency playbook
- Brett Williams (Designjoy): $1.5M ARR productized-agency case study cited in Independent market research Agencies report
- Eric Siu (Single Grain): generative-AI agency operator
- NoGood: AI-augmented growth marketing agency
- ActionPark Media (Victory the Podcast): documented Veritone Voice Network localization case study
Predictions for 2026-2027
- Posthumous and licensed voice deals spread from premium documentary work (Anthony Bourdain, Andy Warhol Diaries) to mainstream advertising and audiobook production by 2027.
- Localization-as-a-service becomes the main agency revenue line, with multi-language podcast and course conversions priced at $500-$5,000 per hour of source audio.
- Through 2027, "digital twin" deployments (cloned voice + avatar for course lessons and personalized messages) become a standard upsell, mirroring the Mondelez / Shah Rukh Khan and Raymond Realty examples.
- A high-profile voice-cloning fraud case forces explicit consent and watermarking standards, which gives an opening to agencies that build provenance into every deliverable.
- Content-volume operators (Play.ht's Podcast.ai-style channels) push voice synthesis into commodity tiers, so the value you add moves toward creative direction and licensing and away from raw production.
Read those predictions as a clock. Every month, more creators decide who handles their audio, and once they settle on a voice and a workflow they rarely switch. The agencies that earn trust this year keep those clients when the tools get cheaper for everyone.
Emerging Opportunities
Podcast localization productized service: Independent market research's Voice Cloning report cites Cherie Hu's audio newsletter and ActionPark Media's localization work. Packaging "your podcast in Spanish, Portuguese, German" at $500-$2,000 per episode gives you a clean recurring offer.
Course-narration and digital-twin services: Berlitz built 8 virtual teachers; Mondelez built a Shah Rukh Khan ad twin. Mid-market course creators and small brands want the same outcome at $1K-$10K project pricing.
Audiobook conversion for KDP authors: Indie authors releasing on KDP rarely produce audio because of the cost. ElevenLabs-narrated audiobooks priced at $200-$800 per book give you a high-volume category full of repeat buyers. How many self-published authors could you name who still have no audiobook?
Licensed-voice rights brokerage: As consent and licensing rules harden, an agency that pre-clears voice talent and resells production rights becomes the trusted middle layer. The Resemble + Andy Warhol Netflix precedent shows how the deal is structured.
Common Objections & Counterarguments
"Voice cloning is creepy.": Consent-first workflows (the Podcastle 70-sentence in-app capture is the canonical example) and explicit licensing make your service feel the same as hiring a voice actor for a series. The discomfort comes from careless defaults, and you can fix those.
"Open-source models will eat agencies.": Independent market research's wrapper-economics counter applies here: distribution, niche tuning and post-production polish compound into a moat. Two agencies running the same models still deliver very different work.
"Voice-over actors will sue.": That is a real risk, with active litigation in 2025-2026. Your defensible position is licensed talent (Resemble, Veritone) and pre-cleared voices, and staying well away from unauthorized cloning.
"AI voices sound robotic.": True for early models, false for ElevenLabs v2 and equivalents on conversational content. The remaining gap is in expressive performance, which is exactly where your human direction still adds value.
Get consent in writing, every time
Everything above assumes the voice you are cloning is yours to use. That never happens on its own, and this section is where you get it in writing.
What separates an agency that can sell to serious clients from one that cannot is a clean chain of permission for every voice it produces.
A synthesised voice built from someone's recordings comes from that person, and in most places you will work the law treats identity as property. Right of publicity statutes, Tennessee's ELVIS Act and its equivalents, and the general law of passing off all point the same way: the voice belongs to the person, and you need their agreement to use it commercially.
Here is what a usable consent document actually spells out, and where most improvised ones fail:
Scope of use. Which projects, which media, which territories. A permission for one explainer video covers that video and stops there. Building a permanent brand voice from the same recordings needs its own permission, and clients often assume they already have it.
Duration. Perpetual or for a fixed term. Perpetual costs more, so quote it as an upgrade instead of assuming it.
Whether the model persists. This is the term people forget. Is the trained voice model deleted when the engagement ends, or kept for future work? The person whose voice it is cares a great deal about the answer, and a contract that says nothing will be read against whoever drafted it. If that is you, how would you feel defending the silence?
Sublicensing and resale. Whether the client may pass the voice on to their own customers, agencies or partners. An open-ended right here means the voice can end up somewhere the original speaker would object to, with your name on the production.
Withdrawal. What happens if the speaker later objects. A clause that admits this can happen and sets a process for it is worth far more than one that pretends it never will.
For a corporate voice identity built on an employee's recordings, add one more: what happens when that employee leaves. Companies commission a brand voice from a staff member and are surprised, a year after the person resigns, to find that the asset they thought they owned is tangled up with someone who no longer works there and no longer wants to be the sound of the brand.
Charge for the permission as well as the production
Most agencies in this space price the work: hours of audio produced, minutes of dubbing, videos delivered. That leaves money on the table, and it gets the risk wrong.
Two projects that take you the same production time can be worth very different amounts. A one-off explainer narration is a small job. A perpetual, sublicensable corporate voice identity that the client will use across every channel for years is an asset with a licence attached. Pricing that by production hours is like billing for a photograph by the shutter click.
You have three levers.
Charge for the scope of rights separately from production. Quote the build, then quote the licence: term, territory, media, exclusivity. Clients already understand this from stock photography and music, so you will not need to explain it, and it turns one number into a negotiation with several parts where only one of them is your time.
Price exclusivity properly. A voice nobody else can use is worth several times one you may reuse. If a client wants the model retired from your catalogue, that is a real cost to you and it belongs on the invoice as a real line.
Charge for retention and maintenance. Keeping a trained voice model available, versioned and re-generatable as tools change is an ongoing service. A modest annual fee for continued access turns a project business into one with recurring revenue, and it is the most straightforward retainer you will find in this niche. How many of your past projects could you go back to with that offer?
What does another month of hesitating cost? One more month of the same paycheck, and one more month where a creator you admire hires somebody else. You already have the tools, or can get them for free tonight. Make one sample, send it to one creator, and let a real reply decide your next step.
Who should skip this
If you are unwilling to handle consent paperwork properly, please do not build voices from real people. The production is the easy half. The permissions are where the liability lives, and an agency that improvises them is selling clients an asset with a defect in it.
If you plan to service outbound calling, get specialist advice before you take the work. The TCPA position above is a summary and should never be mistaken for legal counsel, the damages are per call, and the exposure sits with whoever the regulator decides was making the calls.
If you are hoping the tools alone are the business, you will watch your margins shrink. Your client can run the generation step directly, and it gets cheaper every year. What clients cannot easily do themselves is direction, quality judgement, rights management and knowing when a synthetic voice is the wrong choice.
If the ethics make you uneasy, choose a different niche instead of a different client. This work means reproducing people's voices, sometimes after they can no longer really consent, and if you have not settled your own position on that, you will end up taking a job you regret. Where do you stand on it today?
The uncomfortable question in this business is why a client pays you when they could open the same tool themselves. They already have access. What they lack is four things you can offer.
Direction. Synthetic voice is a performance, and a performance needs someone deciding pace, emphasis, where a sentence breathes and where a line lands flat. Clients who generate their own audio end up with readings that are technically correct and sound wrong, and they usually cannot say why. Naming why is your skill.
Rejection. Knowing which take is not good enough, and being willing to regenerate twenty times to get one that is. A client working alone accepts the third attempt, because they have nothing to compare it against and no appetite for the work.
Rights management. The consent chain described above is administrative and unglamorous, and it is the part clients most want to hand to someone else, especially once their legal team has started asking questions about it.
Judgement about when not to use it. The most valuable advice you can give a client is that a real voice is the better choice for this particular piece. It costs you that job, and it is what brings you the next three, because a supplier who sometimes says no is a supplier whose yes means something.
Everything on that list grows more valuable as generation gets cheaper. It may feel the other way round while you watch the tools improve, so hold on to that.