The Isenzo 7P Method: a structured guide to decide well before implementing.
Most SMBs that rush to deploy an AI customer service agent don't fail because of the technology. They fail because they start with the tool instead of starting with the problem. They pick a chatbot, plug it into the website and, three months later, switch it off because "it doesn't work".
The technology does work. What didn't work was the prior decision-making process.
At Isenzo-IA we have spent months working with consultancies, law firms and clinics that want to roll out an AI agent. The projects that succeed almost always follow the same sequence. We have organised it into seven steps. We call it the Isenzo 7P Method. This guide is not a technical tutorial: it is a decision framework so that, as the person in charge of an SMB, you know what questions to ask, in what order, and when it's time to bring in a specialist.
The Isenzo 7P Method at a glance
Before diving into each step, here's the full sequence. The golden rule: don't skip any. Each step depends on the previous one.
- Purpose: what specific problem does it solve?
- Perimeter: what's in scope and what stays out?
- Personality: how should your agent sound?
- Pillar of knowledge: what information does it draw on?
- Points of contact: where does the agent live?
- Proofing: how do you know it works before launching?
- Polishing: how does it improve over time?
Step 1. Purpose: define the problem, not the solution
This is where most teams go wrong. The right question isn't "do I want a chatbot?". The right question is: "what repetitive task is consuming my team's time and could be absorbed by an agent?".
For a consultancy it might be answering frequent questions about tax deadlines. For a dental clinic, handling appointment requests outside business hours. For a law firm, doing a first triage of inquiries and routing them to the right area.
- Which inquiry or task do you receive more than 20 times a week?
- How many minutes of your team's time does each one consume?
- What percentage of those inquiries have a more or less standardised answer?
If the three answers are solid, you have a use case. If not, you don't need an AI agent: you need to fix another process first.
Step 2. Perimeter: what the agent must NOT do
An agent without a perimeter is an agent that gets your business into trouble. Before thinking about what it will do, define what it won't do. Examples of reasonable limits for an SMB:
- It does not give personalised legal, tax or medical advice.
- It does not promise deadlines or prices without human confirmation.
- It does not handle formal complaints — it routes them.
- It does not access sensitive customer data without verification.
Defining the perimeter has a hidden benefit: it brutally simplifies the project. The narrower the agent's scope, the better it will work and the less it will cost to maintain.
An agent that does three things well is infinitely more valuable than one that tries to do thirty and does them all average.
Step 3. Personality: your brand also speaks in writing
The agent's personality is not decoration. It's what decides whether your customer perceives the conversation as professional or as a disguised form. No need to overcomplicate it. Define three axes:
- Tone: friendly or formal.
- Density: short and direct answers, or explanatory ones.
- Identity: does the agent have a name? Does it introduce itself as AI or as "[your company]'s assistant"?
A practical recommendation: be honest about the agent's nature. Telling the customer they are talking to your company's automated assistant builds more trust than pretending otherwise. And if the customer figures it out on their own, they'll figure it out at the worst possible moment.
Step 4. Pillar of knowledge: what it's trained on
An AI agent is only as good as the information you give it. And here's an uncomfortable truth: most SMBs don't have their knowledge documented. It lives in the founder's head or in email threads. Before building the agent, take the time to prepare three blocks of information:
- Real FAQs: the 30-50 questions you get asked most, with an agreed answer.
- Key processes: how to book an appointment, how to request a quote, what documentation is needed.
- Edge cases: special situations and who to escalate them to.
If this step feels like too much work, that's exactly the signal that your business needs it. The documentation you prepare for the agent will also serve to onboard new team members.
Step 5. Points of contact: where does the agent live?
WhatsApp, website, email, form, social media. The common mistake is wanting to be on every channel from day one. The recommendation: start with one.
Which one? The channel that meets two conditions: it already receives real volume of inquiries and allows technical integration without friction. For most Spanish SMBs today, that channel is WhatsApp Business or the website chat.
Once it works well on one channel, replicating to the others is relatively straightforward. Doing it the other way around, not so much.
Step 6. Proofing: launching without rehearsal is the problem
This is the phase most teams skip and the one that costs the most. Before putting the agent in contact with real customers, subject it to a battery of controlled tests. Minimum viable testing:
- 50 internal simulated conversations with real questions from your history.
- One week of operation in "silent mode": the agent answers, but a human validates before sending.
- A clear list of "acceptable failure cases" and "critical failure cases".
If at this stage the agent makes serious mistakes, don't launch it. Go back to Step 4 and enrich the knowledge base. Launching an agent that isn't ready damages customer trust, and recovering it costs much more than delaying the go-live by a month.
Step 7. Polishing: the agent isn't finished, it's cultivated
Here lies the difference between an AI project that delivers value for years and one that becomes obsolete in six months. A well-managed AI agent reviews every month:
- Conversations that escalated to a human (why did they fail?).
- New questions that appear and don't yet have an answer.
- Changes in your business: prices, services, hours, regulations.
Without that polishing cycle, any AI agent degrades. With it, it improves every month and becomes a real business asset.
Where do you start?
If you've made it this far, you already have something more valuable than a chatbot up and running: you have a framework to decide well.
Most failed AI projects in SMBs are not due to lack of budget or immature technology. They are due to skipping Steps 1, 2 and 4 and rushing to Step 5. If you do the first three well, the rest is execution work.
Isenzo 7P Method scoping template
We have prepared the template we use internally with clients to apply this method step by step: an editable document covering the key questions of each of the 7Ps, with sector examples for consultancies, clinics and law firms. It's free; we ask for your email in exchange so we can send you, once a month, content like this.
If after applying it you find the case is more complex than you thought, at Isenzo-IA we offer practical diagnostics for SMBs from €99. But first, give it a try yourself. You can reach us from the Diagnostic section of the website.
Need help applying the 7P Method to your SMB? Book a €99 Express Diagnostic and we'll help you define the right use case before investing in technology.
Isenzo-IA — Practical AI consultancy for SMBs. Baix Llobregat, Barcelona.
