We present a practical method to eliminate all terms of order a ( a being the lattice spacing) in the calculations of hadronic matrix elements with Wilson fermions. The method is based on an “improved” Wilson action combined with a suitable modification of the fermion operators. With this technique one expects to reduce significantly the systematic errors which arise due to the finiteness of the lattice spacing. We estimate that these errors will be reduced from about 20–30% (as found in recent numerical simulations) to just a few percent.