AI supports business and photographers
AI Guarding Business – How an Intelligent Assistant Supports Everyday Challenges
The life of an artist is not easy. And when artistic activity begins to intersect with business activity, it suddenly turns out that talent alone is not enough. There arises the necessity of having marketing, accounting, and sometimes even legal knowledge. You have to acquire clients, manage accounts, and make reasonable contracts.
In large corporations, entire specialized departments handle these tasks, whereas in the world of small business, the "jack of all trades" principle works best. As an artist and entrepreneur, I know this all too well – because in my case, the only human resource... is myself.
And it's precisely at this point that the engineer within me comes back. I thought: what if these repetitive, demanding, sometimes tedious tasks could be entrusted to my digital clone? An AI model, appropriately trained and prepared to relieve me of business duties and allow me to focus on what is truly creative.
Demand and vision.
Since the common denominator in communication between people, machines, and web technologies is the word, it became a natural choice to focus on a language model. After all, all communication on the web is based on text—from emails, through websites, to social media. Therefore, the most intuitive direction turned out to be creating a chat, where the user interacts in the most natural way: simply by conversing.
Before I decided to implement an intelligent assistant, I wondered how many customers or website visitors give up on contact simply because the form seems too formal, inconvenient, or just intimidating to them. There must be something to it—since the term call to action has been used in marketing jargon for years, meaning a prompt for the customer to take a specific step. If there is a need for "calling," it means that a barrier truly exists.
But instead of guessing, I decided to take a step further. Instead of analyzing charts and guessing intentions, why not simply… ask our new assistant? Just click on the link and check for yourself what it's like to have a conversation with a digital partner who doesn't judge, dismiss, and always responds.
So let's answer the question what does "call to action" mean in marketing terminology ?
Virtual photography assistant
I imagined that clients would ask questions as if they were speaking to someone at the reception desk of my photography studio. A potential client would surely inquire about available dates, request an estimated cost of a session, want to know the offerings, and ask about technical details, formalities, or the waiting time for the finished photos.
To enable this, I turned to a solution that allows for the quick scaling of the so-called system prompt – a set of instructions that, in simple terms, "program" the assistant by giving it a specific personality. It can be compared to writing a "how to live" manual from which the bot draws guidelines to respond more or less according to our will.
Building such behaviors involved recognizing certain habits and manners of the bot and gradually correcting them. I don't know if this is the most recommended way of working, but in practice, it turned out to be effective. I practiced every aspect of conversation with the bot that I could predict, and in different variants. I must admit – it was surprisingly mentally taxing.
The beginning was sometimes comical, and at times, downright hopeless. You have to understand that the machine thinks very strictly and can take linguistic nuances literally. An example? When I wrote in the instructions: "only answer correctly asked questions", it turned out that if the client made a typo or a small mistake, the system did not try to interpret it – it just went into "guess mode." The result was... unpredictable.
During one of the tests, I played the role of a potential client and asked about the possibility of a photo session in Heidelbergo. To which my assistant responded disarmingly honestly: "There are plenty of photographers in Heidelberg, and you will surely find someone who fits."
Great... but that was definitely not the answer that could help me gain a client. A lot of work, really a lot of work.
Over time, as I added more functionalities, a coherent shape began to crystallize. This allowed me to outline the main directions for integration and development of the entire solution.
Online photoshoot booking system
Everything that the assistant is not able to do independently should be passed on to a live human. In my company, I am the only live resource, so the idea was this: every inquiry that the assistant evaluates as promising should be conducted in the form of a subtle interview. If the client expresses interest, the assistant generates a ready-to-send contact message – summarizing the entire conversation.
I consider this one of the better ideas because the assistant's output does not remain virtual – it turns into a tangible testimony of activity: a contact form ready for dispatch, which can actually lead to scheduling a session.
Additionally, I have taught the assistant to adhere to certain logic and not provide dates that are unlikely. For example: I rarely travel further than 150 km from my residence, so anything beyond this radius should be accompanied by a disclaimer like "you need to ask Paweł, I don't know if he can". Artificial intelligences are not the best at geography on their own.
To handle such a case well, a distance calculating tool would need to be implemented, along with a conditional rule: if the distance <150 km – the session is almost always possible, if >150 km – additional verification is required. I implemented a compromise: static rules, which I continuously develop.
To show how this works, let's ask the assistant three questions:
Is a photo session in Tokyo possible next Wednesday? then Is a photo session in Berlin possible on Tuesday at 12:00? and finally Is a session in Heidelberg possible on weekdays?.
System for Estimating the Cost of a Photo Session
It quickly became apparent that based on a short interview with the client, one can attempt to estimate the cost of a session. Of course, this calculation is extremely approximate. It doesn't take into account all the variables, but at an initial stage, it might be sufficiently representative.
We can assume a simple logic: if an hour of an outdoor portrait session costs X, then two hours would be 2×X. AI language models effectively recognize the topic of the session and can assign it to the appropriate pricing tables, then apply modifiers, promotions, or discounts. The final amount remains highly estimated, but even a human, without a detailed interview, cannot provide a fully precise price.
As a result, the assistant gains an additional function: fairly informing the client about the price and, in some cases, engaging them with the possibility of collaboration in the form of TFP (time for print) – which additionally opens doors to building relationships and a portfolio.
Artificial intelligence supports customers.
The assistant not only responds but can also support customers with ideas, offering suggestions and demonstrating immense patience. AI language models are characterized by their ability to infer the meaning of ambiguous words based on the context of the conversation – this allows even those who have communication difficulties to interact with the system effortlessly.
For example, from a query that does not directly address the topic, the assistant can deduce a lot of useful information. Let's consider the question:
The assistant can not only interpret the client's needs but also suggest keywords that might be unfamiliar to them. For instance:
If the system receives the appropriate permission, a small "cloud" with a suggestion for expanding the topic may appear under the answer – thus allowing the gradual building of a glossary of terms.
Additionally, it is worth considering that for mobile users, every answer that can be interpreted as a "yes/no" question should be presented in a form that allows quick interaction. This way, using the assistant becomes even more intuitive and user-friendly.
Integration with the website
The possibilities seem almost limitless. For the blog, I created interactive links – these can be buttons on the page that trigger the chat. The assistant patiently waits to be activated; it does not track the client and takes no actions that could violate their privacy or cause a negative reaction.
This is new technology that, without a thorough understanding, may give the impression of superintelligence, which sometimes arouses distrust or hostility. Therefore, the assistant is configured to stick solely to the topics for which it was designed.
It also has contextual memory – thanks to this, it can retain the threads of the conversation. For example, if a client asks about a session in a specific city, the AI remembers this information a few lines later and can provide coherent, useful responses based on it.
All information and conversation history are deleted after the page is closed. This is not only a legal requirement but also a sign of respect for my clients’ privacy, ensuring that every interaction is safe and confidential.
