Hi Denys, thank you for your interest. In general I do understand why big corporations need to be implementing AI to improve efficiencies and/or customer satisfaction, without questioning too much the ratio cost/benefit. However, for small SMEs cost comes first, and determining when to invest in an AI solution gets tricky. Today I have the impression that developing in AI is costly. Could you talk about my perception that AI is costly for a SME?
sorry for a late reply - I missed the notification!
I agree that for SMEs the cost comes first, this is why I always say “show me the money!”, meaning the additional revenue or cost savings that can be achieved with AI. I’ll talk about it in today’s webinar in more detail. This allows for a clear understanding of ROI and thus justifies a decision to implement AI.
On the side of pure cost, I wouldn’t say that AI is more expensive than a normal software project. Due to the experimental nature of AI, it might take longer until you arrive at a working solution (since you need to train and retrain the models etc). However, many of the most advanced AI tools are free (frameworks like TensorFlow). Another part, such as deployment & hosting, is more or less the same as for a normal software solution. Compute power on the cloud can cost more since again, the ML models need to be trained.
Hope it helps! Join our webinar to know more about the business case :)
(1) A B2B client found that an “AI-enhanced” approach worked better and allowed customer service agents to be more productive and provide faster and better help to customers. AI provided sentiment analysis, recommended answers and next best action suggestions for agents to work with, but all direct customer interactions remained human. It seems like a much lower-risk approach, potentially faster to implement and with easier to measure impact, than striving for a fully autonomous chatbot. Could you talk about differences in approaches and how different implementation processes compare? (And potentially build on one another)
(2) Do you have any tips, tricks and warnings for multi-lingual AI implementations? What can and can’t be shared between the different language tracks?
This sounds like a chatbot heavy question - so I’ll try to answer to the best of my ability.
1) Human in the loop systems are very popular. It’s how the chatbot is trained, and also how the human touch & quality is preserved.
You are descriving an even less “chatbotty” implementation, which is very reasonable in some cases. The problem is, it won’t help you with 24h availability & immediate responses. So it’s more of a business than a technical decision.
2) Here I can’t say too much unfortunately - I only know that most NLP systems work best for English. HuggingFace work to solve this problem: https://huggingface.co/models
This will mark this comment as best reply and close your question.
Are you sure?
This will close your question without a Best reply.
Are you sure?
This will report this content as inappropiate to the moderators.
Are you sure?