AI for Business Owners: Where It Delivers ROI and Where It Does Not
- Lior Izik
- Business Intelligence & AI
- October 15 2026
By Amazing Business Results | Certified Zoho Premium Partner
A straight conversation between Zoho AI specialist Zach Zivnuska and ABR’s Lior Izik — practical AI, Zoho CRM, agents, and where to actually start.
Topics: AI ROI · AI Agents · Zoho Zia · AI Adoption · Business Strategy
Most AI content right now is noise. Business owners are tired of being told AI will change everything while they watch thousands of dollars leave the account every month with nothing to show for it.
So this was a different kind of session: the platform and the implementer, talking straight. Zach Zivnuska works on AI at Zoho. Lior Izik runs Amazing Business Results, the most reviewed Zoho Partner in the world. No pitch, no hype — just where AI pays off, where it quietly burns money, and what a business owner should actually do first.
Where AI Actually Delivers ROI
The value sits at the bottom of the pyramid, not the top. The strongest returns come from frontline, high-volume busy work — the swivel-chair tasks: answering tickets, responding to calls, sales outreach that often gets no reply. The work is quick, it needs some knowledge, and frequently nothing comes back from it. AI can package that response or outreach almost instantly, with no human in the loop. That is real ROI.
Go up the ladder and it collapses. Middle management and executives live in an inbox. There is often no clean process, no examples, no structured data to point an agent at, and you are not going to turn an agent loose inside someone’s email. At that tier the return is marginal, often around 4 to 6%. If a leader is spending hours a day just making AI useful for themselves, the honest question is whether you would be better off hiring a person to do the work instead.
| Where the task sits | AI ROI | Typical tasks |
|---|---|---|
| Frontline, high volume | High | Support, booking, auto-reply, outreach |
| Middle management | Low | Lives in the inbox, little clean data |
| Executive, judgment work | ~4–6% | No fixed process to automate |
The pattern
The more repetitive and high-volume the task, the higher the ROI. Support, appointment booking, auto-replies — strong. Executive judgment work — weak.
Two of the strongest frontline cases are already written up in detail: an AI receptionist → that answers the calls your team cannot reach, and a tier-one support agent drafting replies from your own knowledge base. NOTE: verify the shared inbox post URL once published.
Why the Same AI Produces Greatness and Garbage
Same models, wildly different results. The difference is not the tool, it is the use case and the discipline.
Most failures come from throwing spaghetti at the wall: trying twenty things at once, seeing one stick, and declaring AI a failure. The businesses that win do the opposite. They pick one use case, apply it, measure it, and get a real result. They may not use AI anywhere else. But that one thing works, and they can prove it.
The Two Things That Quietly Sink a Project
The wrong question. A client says “email me every new lead.” You ask why. “Because a new lead came in.” How many a day? Thirty. So you want thirty emails? The real question was never “how do I get an email” — it was “what are my options?” Maybe it is an auto-updating view. Maybe a mobile push. Feed the wrong question in and AI will efficiently hand you back the exact mess you were trying to escape, in a nicer interface.
Contaminated data. A client insists the AI report is wrong: “I only got six leads this month, not thirty.” They did get thirty. Six were real; the other twenty-four were test records they ran last month. AI did not fail. The data did.
Finding contaminated data before it reaches an AI model is one of the first things a Zoho Platform Assessment looks for — duplicates, test records and half-empty fields are exactly what turns a confident AI answer into a wrong one.
The first problem is harder to see from the inside. It is also why the process question matters more than the tool question: where a process has never been written down, as described in the playbook approach, there is nothing for an agent to automate reliably.
What Zoho Zia Does Out of the Box
You do not always need developers. Zia is built into each Zoho app at the feature level, tuned to what that app actually does — and it does not punch you in the face.
In CRM that is anomaly detection, workflow suggestions, and built-in SDR agents to help manage a deal. In Desk it is a tier-one agent answering fundamental questions from your knowledge base and past tickets. It elevates the experience you already have rather than bolting on something alien.
When your situation is genuinely custom, you take a foundational Zia agent — like that tier-one support agent — and craft it around your own SOPs and knowledge. That is the point where a partner earns their keep.
One Person, Twenty Agents
Here is where it stops looking incremental. One employee can now run twenty agents, each qualified at the level of a basic employee, working around the clock. One person, replicated into twenty, 24 hours a day. That is not a productivity tweak.
And the internet comparison is both wrong and right. The internet made information accessible — you could look something up, learn, and act. AI takes information and turns it into actionable knowledge, then lets the machine act on it. If you remember how fast the internet compounded from 2005 to 2012, expect something similar here, except this time it is action, not just answers.
Spending Thousands on AI With No ROI
“My team loves AI and we are getting great results, but no new money is coming in.” Common, and usually fixable. The trap is pouring budget into back-office efficiency on roles that do not generate revenue. Getting more efficient at work that does not bring money in still does not bring money in.
Two moves.
First, measure honestly. A leader running a tool like Claude Code to generate and validate reports for roughly a thousand dollars a month may be covering what a full salary plus benefits used to cost. Expensive per month, cheap per year.
Second, when AI frees someone up, move them toward revenue. One medical-testing company trimmed an intake team from four people to two, then asked the freed-up staff whether they wanted to retrain as lab techs, the role that actually generates revenue. Idle capacity is not a layoff sentence. It is a redeployment opportunity.
Lior’s own commitment, made about three years ago: no employee gets laid off because of AI. The team trained weekly, in parallel, on non-live projects. It was painful — people resist change — but today AI is simply part of the work, and roles were restructured so one architecture baseline flows to developers and QA alike.
The honest part
Even in 2026 the ROI is not always clean. A human still has to check everything AI produces, and mistakes happen. But the alternative is worse: a business that refuses AI gets outrun by competitors who adopt it. It is not optional anymore.
Where to Start if You Are a Business Owner
Keep it boring and disciplined.
Pick a provider and commit. Do not hop between tools for six months. You will see little and win less.
Start at the most fundamental layer. Sales and marketing — high-volume auto-replies, lead generation, email content tailored to the customer profile — is the lowest-hanging fruit. No human action required.
Take one small piece at a time. Solve it, measure it, then move to the next. Start in the middle or at the executive level and you will just confuse yourself.
Do not spread yourself thin. Launching across sales, support, implementation and everything at once wastes budget, kills focus, and hands resistant employees proof that “it did not work.”
And accept the possible outcomes: it works, it does not, or you discover you did not need the process at all. All three are fine, because you found out cheaply. That is the same milestone logic ABR uses on implementations. Nobody rolls out 55 applications at once. You get one or two right, bank the win, then build the next.
If you want that first use case chosen with evidence rather than instinct, our AI Readiness Assessment does exactly this: it ranks candidate use cases by value against effort, lists what has to be fixed first, and tells you honestly if the answer is “not yet”. NOTE: verify this URL once the page is published.
About Amazing Business Results
ABR is the most reviewed Zoho Partner in the world, with 2,000+ completed projects and over 12 years of experience across the US, Canada, and Europe. Our team of 30+ Zoho-certified professionals extracts your real business requirements and translates them into custom Zoho systems, backed by comprehensive development and training so your people actually use them the right way.
Ready to Put AI to Work Where It Actually Pays Off?
Explore our related services:
The Top 20 CRM Automations in the World covers twenty systems with real client stories and the exact builds behind each one.
Start your free Zoho One trial (no credit card required): USA | Canada | Europe
Subscribe to the ABR Dispatch — Free Weekly AI and CRM Newsletter: zfrmz.com/54B40ZnE2NSRhC3s1Yoh
Top Zoho Extensions: ABR on Zoho Marketplace
Zoho CRM Course: amazingbusinessresults.com/zoho-crm-course