The boss has been "brainwashed" by AI vendors, how can the IT director survive?

"Your IT team is just so conservative! The vendor already said the AI can go live in three weeks, why do you say it takes three months?"

"It worked perfectly during their demo, why did it stop functioning once it's in your hands? Is there something wrong with your capabilities?"

"Don't talk to me about data governance or process sorting, those are all excuses. I'm just asking you, can you do it or not?"

If you are an IT leader in the AI pilot phase, don't the above remarks sound familiar to you?

The boss who says these things is not stupid. He was just "brainwashed" by the sales pitches of some vendors. With carefully packaged demos, selective success stories, and concept-swapping technical rhetoric, vendors planted in his mind the illusion that "AI can do anything". As a result, under this illusion, the following problems arise: the professional judgment of the IT team becomes conservatism, historical accumulation becomes a burden, and basic common sense becomes nonsense that hinders progress.

At this point, the IT department will face such a problem: You understand the technology, but the boss doesn't trust you; the vendor may not understand your business, but the boss trusts them.

I. Why does the boss get "brainwashed"?

Don't rush to call the boss stupid in your mind. Only by understanding why he was brainwashed can you find a breakthrough.

1. The "lifeline" mentality driven by anxiety.

Under the current economic environment, most traditional enterprises are facing the pressure of growth and survival, and bosses are eager for a starting point that can turn the tide. At this time, AI is just packaged as that lifeline. Vendor sales know this well: what they sell is not technology, but hope. When a person is extremely eager for a certain result, he will actively filter out the unpleasant information.

2. "Authority transfer" brought about by information asymmetry.

The boss does not have a technical background, so he cannot judge whether what the vendor says is true or false. But he can see that the flow chart on the PPT is very beautiful, the demo is very smooth, and the names of the case companies are very impressive. He will also think: They have served so many big customers, they would not lie to me, right? This information asymmetry makes the boss outsource the technical judgment right to the vendor.

3. Aesthetic fatigue of internal staff.

The IT team reports to the boss every day that "the data is not clean", "the process needs to be optimized", and "it takes time". After hearing this for a long time, the boss will suspect that you are making excuses. But vendor sales always say "no problem", "it can be done", "we will help you get it done". One makes you feel comfortable, the other makes you feel bad, and human nature naturally tends to trust the one that makes you feel comfortable.

It is not difficult to see from the above that this is not a problem with the boss's IQ, but the default setting of human nature. So it's useless to call the boss stupid in your mind, and complaining is also useless. You have to design your response in line with this human nature.

II. The Dilemma of IT Leaders

In this situation, IT leaders are not only facing technical challenges, but also survival challenges.

Dilemma 1: Telling the truth is regarded as conservatism, while telling a lie goes against your own principles.

You know clearly that the vendor's promise of "going live in three weeks" is impossible, but when you say "it takes three months", the boss thinks you are evading responsibility. If you bite the bullet and say "okay, let's try", once the project fails, you will still be the one to bear the consequences.

Dilemma 2: Historical accumulation is regarded as negative assets.

The data standards, process specifications and system architecture you spent five years building are regarded as old antiques by the boss. When the vendor says "you don't need to worry about those, we have AI", the boss will think that all your previous work was in vain. Your professional accumulation is worthless in front of "new concepts".

Dilemma 3: Caught between the vendor's promises and the boss's expectations.

Vendor sales pat their chests in front of your boss, but cannot deliver on various points once the project starts. When you follow up with the vendor, they say your data is too poor and your business is too complex. When you explain to the boss, the boss says the vendor has already pointed out that the problem is on your side. At this moment, you are stuck in the middle and get blamed from both sides.

Dilemma 4: Once the pilot fails, you become the sole person in charge.

The vendor can leave without any burden and continue selling to another customer. The boss will not admit that he has been fooled, he will only think that the IT team has poor execution. You become the only person who can be held accountable in the entire decision-making chain.

The superposition of these four dilemmas has become the real situation of many IT leaders: You are using your expertise to shield the company from risks, but every time you successfully block risks, it becomes evidence that you are not cooperating.

III. Survival Strategies

Lao Yang believes that the core principle at this moment is only one: Do not try to convince the boss with technical logic, use facts and risks to make the boss wake up on his own.

Strategy 1: Transform "objection" into "conditions".

Don't say "this can't be done", say "it can be done, but the following conditions need to be met".

When the boss relays the vendor's promise, your response can be like this: The solution mentioned by the vendor does have prospects. After careful review, to implement it, we need to solve three preconditions first. First, customer data needs to be cleaned, which will take about X weeks. Second, we need to interconnect the three systems A, B and C, which requires coordinating X departments. Third, at least X weeks of trial operation and tuning are required after launch. We can work with the vendor to write these three conditions into the project plan, clarifying who is responsible for what and when it will be completed. If all conditions are met, we promise to go live on schedule.

The clever part of this statement is: you do not deny the vendor, nor do you deny the boss, you turn "can't do it" into "what conditions are needed". At the same time, you also pull the vendor into the responsibility circle. If the conditions are not met, it is not your problem, but the preconditions are not in place.

Strategy 2: Replace "big promises" with "small pilots".

Don't argue with the boss about "whether AI works" when he is at his most enthusiastic. Use a minimum verification with extremely low cost to replace a debate that is doomed to have no conclusion. When the boss says "use AI to build an intelligent customer service for the whole company", you can say: Boss, let's spend two weeks first, use real customer service data to do a small verification, and see what the accuracy rate can reach. If it reaches 80%, we will roll it out comprehensively; if not, we will adjust the plan. Two weeks later, show him the results generated by running the real data. Facts speak louder than words.

Strategy 3: Translate "technical risks" into "business language".

Bosses can't understand "data consistency", "interface coupling", "technical debt", but they can understand "customer complaints", "financial reconciliation errors", "business outage". Don't say to the boss "if the data is not interconnected, AI can't work". Say "if the data is not interconnected, the customer information given by AI may be wrong, and salespeople using wrong information to communicate with customers may lose orders". By speaking out the business consequences, the boss will really pay attention to it.

Strategy 4: Establish a "vendor evaluation file".

Every time a vendor comes for a demo, arrange a special person to record: what promises they made, what the demo environment is, whether the customer cases are real, and whether they can provide POC afterwards. Establish an internal evaluation form to record the gap between each vendor's promise and actual delivery. When the vendor comes to make empty promises again, you can show the data: Boss, the last vendor A promised to go live in three weeks, but it actually took three months, and the delivery quality also had problems. For vendor B's promises this time, we suggest doing POC verification first. Speaking with historical data is more powerful than any technical demonstration.

IV. Deeper-level Construction

The above are survival strategies to solve the immediate problems. In the long run, enterprises also need to establish a set of mechanisms to ensure that the professional judgment of the IT team can be truly heard.

Construction 1: Establish an "AI Project Evaluation Committee".

Set up an evaluation committee composed of IT, business, finance and procurement, and form a set of scientific and feasible selection mechanisms. All AI projects must go through technical feasibility assessment, input-output calculation, and risk assessment. Vendor demos must be presented in front of the entire committee, not just to the boss. This measure can effectively filter out sales pitches that only make empty promises to the boss.

Construction 2: Incorporate "vendor promises" into contract terms.

Don't let the vendor's verbal promises only stay in the demo stage. Write key promises into the contract, and set technical acceptance indicators and compensation clauses for non-compliance. For example, promise that the accuracy rate of intelligent customer service is not less than 85%, and if it fails to meet the standard, the project payment will be deducted proportionally. With contract constraints, vendors will be much more cautious when making promises.

Construction 3: Cultivate the boss's "technical judgment".

This point is the most difficult, but the most fundamental. You can regularly share "AI truth" with the boss, not about technology, but about real cases. For example: Boss, that vendor we met last time was held accountable by their customer later, and the reason is like this. Use real industry cases to help the boss build rational expectations for AI.

Construction 4: Accumulate trust with "small victories".

Don't expect the boss to trust the IT team from the very beginning. Use a series of small victories to accumulate trust. For each pilot, choose a low-risk, fast-effect scenario to produce quantifiable results. When the boss sees that the IT team can really get things done, his trust in IT will increase, and his superstition in vendors will decrease.

V. Final Summary

It is not difficult to see from the above that when IT leaders are in the situation of "the boss is brainwashed by vendors", there are two extremes that should be most avoided. One is to confront head-on, say that this is impossible, and be marginalized by the boss as a conservative. The other is to follow blindly, say okay and listen to the vendor completely, and you will still be the one who bears the consequences.

The correct approach is: do not deny the boss's expectations, do not deny the vendor's solution, but break down the expectations and solutions into verifiable conditions and measurable risks. Use facts instead of arguments, use pilots instead of promises, use data instead of feelings.

You cannot change the reality that the boss has been brainwashed by vendors, but you can let him wake up on his own after seeing the facts. This process requires patience and strategy. The core of the strategy is actually only one point: speak with facts, prove with data, and accumulate trust with small victories.

This article is from WeChat Official Account "Xiangjiang Digital Review" (ID: benpaoshuzi), author: Lao Yang, authorized for release by 36Kr.